Installation

Before you can create your first optimization model with aristopy you need to install the following three essential components:

  1. Python and IDE

  2. Aristopy package

  3. Solver

Python and IDE

The first requirement for using aristopy is a working installation of Python on your machine. Aristopy is currently tested with Python 3.6 and 3.7. Please use one of the many good tutorials on the internet if you need help with the Python installation. Make sure to add the path of the installation to your system’s environment variables to simplify the call of python, pip, etc. from the command line. To enhance your workflow with Python, we also recommend installing an integrated development environment (IDE). PyCharm has served well for us, but there are numerous other useful and free software tools available.

Aristopy package

You can easily install aristopy in your current environment via pip by using the following command:

>> pip install aristopy

Alternatively, you can create a clone of aristopy’s GitHub repository (provided git is installed) in a local directory of your machine

>> git clone https://git.tu-berlin.de/etus/public/aristopy.git

or download a zipped version directly from the GitLab page.

After that, you need to go to your local directory and install aristopy by running the setup-file with python

>> python setup.py install

or using pip install 1.

>> pip install -e .[dev]

Solver

Note

The installation of aristopy does not include any solvers. They need to be obtained separately, in accordance with the properties of your model, the availability of licenses, and your specific preferences.

The availability of a mathematical solver is essential to generate results for your optimization problem. You need to ensure the solver of choice is suitable for your model’s mathematical class (e.g., if you added non-linear correlations, you need to use a solver for non-linear problems). We refer to the common literature for further information on mathematical modeling in general.

There is a great variety of solvers available on the market. For the use with aristopy you have to consider that the solver interface must be usable for the underlying optimization package Pyomo. A useful way to start is downloading the open-source solvers, available for free from AMPL. For example, we recommend the solver CBC for mixed-integer problems (MILP) and the solver ipopt for non-linear (NLP) problems.

If you have access to a license of the powerful, commercial MILP-solvers Gurobi or CPLEX, you are encouraged to apply them to solve your aristopy model (provided your model is not non-linear). Academic users may be eligible to receive free licenses for both solvers. Please note that all installed solvers must be locatable by aristopy. Therefore, it is important to add the path to the solver executables to your system’s environment variables.

1

Argument -e is optional for editable / development mode. Add [dev] if you also want to install the extra-dependencies, i.e. sphinx, pytest, etc.