melloflavio / 2019-MSc_Thesis

Cost optimization in frequency control of unbalanced distribution systems

This repository was developed as portion of the final thesis of the MSc Data Science programme at City, University of London. The final report on the work developed can be seen at ./report

Version: 1.0

Prerequisites

• PyEnv: Ensures correct python version (as per .python-version)
• PipEnv: Ensures dependencies & locks down package versions

Running Code

1. Starting with a system with python v3.7.3 and the corresponding pip version installed Note: Pyenv and pipenv are optional, but facilitate the process of ensuring the correct versions of the libraries
   1. If using pipenv, ensure you are within the virtualenv: pipenv shell or issue commands with pipenv run <command>
2. Install all libraries specified in section B.2. Ensure the correct versions are being installed.
   1. If using pipenv this can be done running the command pipenv install, which automatically installs all the libraries as specified in the pipfile
   2. If relying on pip only, each library can be installed individually
3. Start the jupyter notebook server: (jupyter notebook)
4. Run the experiments found in the form of notebooks in the ./app/experiments folder
5. A template notebook can also be found which provide the blueprint for running future experiments. ./app/experiments/Template-Experiment.ipynb

Troubleshooting

Jupyter Notebook

Problem: Jupyter Notebook fails to run. Usually indicating an issue of invalid syntax pointing to typing hints. Fix: Ensure the python kernel running in jupyter is of the correct python version (tested with 3.7.3). To show version in any notebook:

    from platform import python_version
    print(python_version())

To update the kernel in the jupyter installation to your currently installed python version (using pip)

    python3 -m pip install ipykernel
    python3 -m ipykernel install --user

Problem: Modules installed through pipenv are not being found within a jupyter notebook Fix: You need to ensure the kernel created points to the dedicated virtual environment in which pipenv installs the packages. First you need to find out the name given to the virtualenv created by pipenv. In a terminal session, navigate to the project's root and type:

  $ pipenv --venv
  => ~/.local/share/virtualenvs/VIRTUAL_ENV_NAME

With the virtual env name in hand, it is now time to declare a new kernel for jupyter pointing to that specific environment

  pipenv run python3 -m ipykernel install --user --name=VIRTUAL_ENV_NAME

Finally, open the desired notebook and ensure the correct kernel is used using the top menu: Kernel => Change Kernel => VIRTUAL_ENV_NAME

Alternate fix: If nothing else works, as a final alternative one could install all the packages listed in the pipfile using whatever python package manager they prefer (conda, pip, etc). Ensuring the versions match the ones described in the pipfile. This is not recommended, as the preferred way would be to ensure the project is running as initially specified, but it can be an alternative where nothing else seems to work.