Power Kiosk Capstone - Forecasting Energy Usage

Time Series Prediction Pipeline Overview

Set up the environment

If running on MacOS Create and activate conda environment

make conda_create
conda activate power_kiosk
make install

If running on Windows

conda create --name power_kiosk python=3.9 pip
conda activate power_kiosk
conda install -c conda-forge -c pytorch u8darts-all
pip install -r requirements.txt 

Install ODBC Driver

To connect your MSSQL Database with Python follow the instructions listed here

Run the entire pipeline

To run the entire pipeline, type:

1. If running from the base folder

python scripts/run_pipeline.py --cl <contract location id(s)> 

2. If running from scripts folder

python run_pipeline.py --cl <contract location id(s)>

Important details about running the pipeline

The pipeline takes 3 inputs:

• --cl: A single contract location ID, a list of contract location IDs (comma separated, with no space) or the keyword "all" for all contract location IDs.

• --p: Number of periods or months you want the forecast for. Default value is set to be 12, which means the output of the pipeline will give you 12 months forecast into the future for contract location IDs in the input.

• --n: Minimum number of months a contract location needs to have in order for it to be part of the model. Default value is set to be 36 months.

Examples

• Single contract location ID

python scripts/run_pipeline.py --cl 0082c329a35944de939acdfb5975dd23

• Multiple contract location IDs

python scripts/run_pipeline.py --cl 0082c329a35944de939acdfb5975dd23,0219a6756d3e439d84f5bb5678f40499,07e2ba4b87b04684b4ea75c5654d354d,0d674bb909474caeb24cccc0d051df92

**Note there is no space between two contract location IDs in the input.

• All contract location IDs

python scripts/run_pipeline.py --cl all

• Changing the number of periods to forecast from 12 to 5

python scripts/run_pipeline.py --cl 0082c329a35944de939acdfb5975dd23 --p 5

• Changing the minimum number of months required from 36 to 30

python scripts/run_pipeline.py --cl 0082c329a35944de939acdfb5975dd23 --n 30

Run Streamlit to visualize output

To launch the streamlit app locally, run:

streamlit run src/build_streamlit.py