2022-2023

Instructions

The project will be two-fold: handle datasets from French electricity consumption with the motivation to create

  1. a visualization inspired by the image below,
  2. to predict electricity consumption on a future day.

Electric consumption in France

Data link for prediction: for this part the dataset is available at https://odre.opendatasoft.com/explore/dataset/eco2mix-national-tr/information/?disjunctive.nature&sort=-date_heure. The goal is simply to predict the consumption on Thursday 8 December for the whole day, at the country-wise scale. We will measure your score as a Mean Square Error on the 96 quarters of hours during that day, so your output should simply be a .csv file with the following structure:

Date Heure Consommation (MW)
2022-12-08 00:00 30811
2022-12-08 00:15 30889
… … …
2022-12-08 23:45 30878

You will also display your prediction for all the sources (Gas, wind, etc.) on a simple graph of your choice.

Data link for visualization: For this part, the dataset is available at https://data.enedis.fr/explore/dataset/consommation-annuelle-residentielle-par-adresse/information/.

The motivation is to analyze and understand better the electricity consumption of the French household.

The main task for this part of the project is to provide an interactive map of France at the city-wise level displaying the yearly average household electricity consumption between 2018 and 2021. The map should be clickable, and when clicking on a city, you would display the distribution (with a violin plot, a swarmplot or a KDE plot for instance) of consumption per household in this city over the four years.

Hint: you could be inspired by this gif

Could you also spot for Occitanie and Herault which cities have the largest/smallest consumption?

Intermediate tasks

  • data cleaning steps: removing missing values if need be.

Major objectives:

  • displays interactive maps.

Useful packages (non-exhaustive list): - osmnx - plotly - folium - networkx

Timing

  • Mid-term project snapshot: Due date November 15, 23:59. This will include the creation of a github repository (README.md, etc.), a short description of how the work is shared in the team and a detailed work program for the project including the tasks’ attribution (coding).

  • Preliminary presentation: An oral presentation with Beamer (duration: 5mn), in-person on date/place TBC (December 2, 11:30 to 13:00).

  • Due date (final project): The github repository should be completed before Tuesday 6 December (23:59). Nothing pushed after the deadline will be taken into account. The oral presentation (max: 20mn) will be in-person Friday 9 December (8:00AM) (room: 36.4).

Elements expected / Grading

Summary

General Details Points (out of 20)
Mid-term Git / branches 1.5
Task affectation 1
Dataset creation 1
Preliminary-term Beamer (structure, spelling) 0.5
Code Science Technical Problem Resolution 4.5
Readme/Comments/Pep8 1
Unit Tests/CI/Deploy : wheel 1
Class (create at least 1 class) 0.5
Reproducibility/Dataset loading 1
Graphical aspects: Widgets, clickable map, etc. 2.5
Time/Memory efficiency 1
Documentation 1.5
Oral Beamer (structure, spelling) 1
Clarity / lively presentation / Rhythm / Show 2
Total 20

Details

  • The ultimate goal is to provide a Python module that can be imported with pip and contains your work. A description of the procedure will be needed (imagine you are addressing to a user not aware of your package). An example of a project made in 2020, is available at https://github.com/tanglef/chaoseverywhere.

  • Your Python module should have two submodules corresponding to the two main tasks (prediction and visualization)

  • The project must be stored on a github repository.

  • You have to choose a name for your project. Hereafter, it is denoted by my_module_name.

  • It should contain all the aspects described below.

Science

Solve (even partially) the problem raised in your project description.

Project structure

  • All the code will be placed in a subdirectory called /my_module_name (choosing your module name accordingly).

  • A presentation (in an open source format: like Beamer, with TeX, see for instance some templates here https://github.com/josephsalmon/OrganizationFiles, or LibreOffice Impress) will be put in a /beamer directory. The later will be a short presentation of the work that will be orally presented during 15mn in front of a jury.

  • A documentation (using sphinx) will be stored in a /doc subdirectory.

Git aspects

  • A (markdown) readme.md file introducing your work and the team member (first/last name + email).

  • A .gitignore that prevents garbage files to be included in your project.

  • Equilibrated commits in two branches should be done (e.g., in the development branch and the master one), and merged for the milestone day.

Object programming aspects

  • You should code at least one Python class.

Dataset(s)

  • The data used should be available in a way that the end user does not need to perform a manual download of any kind (use the download package or variants for instance).

Graphical aspects

The repository will contain code of the following nature:

  • A code producing a (possibly interactive) map.

  • histograms/kde/swarmplots/etc. plots illustrating the data.

Time/memory evaluation

  • A full study of the time and memory footprint of the code produced will be provided for the whole pipeline used.

Documentation

  • Documentation should be added correctly for the functions written. Please use sphinx.

Test and CI

  • Provide unitary tests to check that the function you proposed satisfies the requirement you target.
  • Implement a Continuous Integration solution with github that runs your unitary test at each commit.

Deploy

  • It should be possible to package your Python module using wheel (i.e., you need to provide a setup.py, file).
Back to top