Energy consumption prediction

Predicting energy consumption based on weather data. To predict energy demand based on weather consumption two models using skforecast will be developed. The goal is to use univariate and multivariate Timeseries forecasting with sklearn-based regression models. Also, a comparison between the uni- and multivariate approach will be done.

Setup

This project uses anaconda for project setup.

Linux & MacOS

To create the virtual environment, please execute the following set of instructions.

conda env create -f ./env.yml
conda activate weather-prediction

Note: If some updates are done to env.yml you will have to update the virtual environment running the following command.

conda env update -f ./env.yml

Notebooks

The repository contains only one Jupyter-Notebook with all relevant steps including EDA, Pre-Processing, Model-Implementation and -Evaluation. The dataset was downloaded from [1] and contains the electrical and heat demand of a residual building from December 2010 until November 2018 in one-hour steps.

References

[1] TAHERI, S., M. JOOSHAKI und M. MOEINI-AGHTAIE, 2021. 8 years of hourly heat and electricity demand for a residential building [online]. IEEE DataPort [Zugriff am: 3. Januar 2023]. Verfügbar unter: doi.org/10.21227/dfvb-re49