Power prediction and analysis in France

Description

A datascience project that shows the power production in France and predicts the consumption.

• A pipeline : to get, transform and serve the data
• A dashboard : to visualize the power production and prediction
• Prediction models based on fbprophet : to predict the power consumption for the next day, week, month or year!

Installation

Two installation methods are povided, one with Python and pyenv, the other one with Docker.

Installation with Python and pyenv

A simple way to install the librairies is to use a virtual environment powered pyenv. This method has been tested with the following versions of Python:

• 3.8.10
• and that's it for now!

The first step is to download the right version of Python according to the list above:

pyenv install -v <python_version>

The next step is the creation of the environment :

pyenv virtualenv <python_version> <project_name>

To activate the freshly created environment :

pyenv activate <project_name>

After the activation, install the requirements :

pip install -r requirements.txt

That's it !

Installation with Docker

In this case, you only need Docker. The installation process will take place during the build of the container. To build it :

docker-compose build

Run the projet

The project can be used in two ways :

• as a pipeline
• as a dashboard
• as a container

INFO: During the first launch, the three datasets are downloaded : this can take some time, approximately 5min ;)

Run with Python

As a pipeline

This method allows you to run the pipeline without the dashboard :

python src/pipeline.py <options>

To see all available options, please use the -h option first.

As a dashboard

The dashboard is launched with the following command. Please look at available options with the -h option first.

python src/app.py <options>

Run with Docker

The container is launched with the following command :

docker-compose up

Open your web browser and go to <your-local-ip>:8050 to see the dashboard. Please adjust the ip address and the port to your needs in the docker-compose.yml file and the Dockerfile file.

Todos

Many things can be done to improve this project. For example :

• add unittests
• get the hourly temperatures from meteostat instead
• provide other prediction methods and compare them
• add holiday periods in France : during Christmas holidays for example, the consumption prediction seems to be too high

Licenses

The datasets are provided under Licence Ouverte / Open Licence V2.0 licenses.