This is the official repository that implements the following paper:
Zhe Wang, Han Li, Tianzhen Hong, Mary Ann Piette. 2021. Predicting City-Scale Daily Electricity Consumption Using Data-Driven Models. Submitted to Advance in Applied Energy
We developed data-driven models to predict city-scale electricity consumption.
- We developed and compared four models: (1) five parameter change-point model, (2) Heating/Cooling Degree Hour model, (3) time series decomposed model implemented by Facebook Prophet, and (4) Gradient Boosting Machine implemented by Microsoft lightGBM.
- We applied our models to explore how extreme weather events (e.g., heat waves) and unexpected public health events (e.g. COVID-19 pandemic) influenced each city’s electricity demand
git clone https://github.com/LBNL-ETA/City-Scale-Electricity-Use-Prediction
cd City-Scale-Electricity-Use-Prediction
Set up the virtual environment with your preferred environment/package manager.
The instruction here is based on conda. (Install conda)
conda create --name cityEleEnv python=3.8 -c conda-forge -f requirements.txt
conda activate cityEleEnv
bin: Runnable programs, including Python scripts and Jupyter Notebooks
data: Raw data, including city-level electricity consumption and weather data
docs: Manuscript submitted version
results: Cleaned-up data, generated figures and tables
You can replicate our experiments, generate figures and tables used in the manuscript using the Jupyter notebooks saved in bin: section3.1 EDA.ipynb, section3.2 linear model.ipynb, section3.3 time-series model.ipynb, section3.4 tabular data model.ipynb, section4.1 model comparison.ipynb, section4.2 heat wave.ipynb, section4.3 convid.ipynb
Notes.
- Official Documentation of Facebook Prophet.
- Official Documentation of Microsoft lightGBM.
Feel free to send any questions/feedback to: Zhe Wang or Tianzhen Hong
If you use our code, please cite us as follows: