expected stdout:
total associated CO2: 20.122 lbs
total fridge run time: 10 days, 8:00:00
Estimated time spent (not including research time): ~10 hours
pip install -r requirements.txt
cd src
python refrigerator_sim.py
docker build -t smart-fridge .
docker run -v /my/local/output/dir:/root/output smart-fridge
For each five minute interval t, the constraints
fridge_on(t) ∈ [0,1]
33 <= fridge_temp(t) <= 43
fridge_temp(t+1) = fridge_temp(t) + 5/12 - (10/12 * fridge_on(t))
fridge_temp(0) = 33
and objective function
[moer(t), ..., moer(t+n)]^T * [fridge_on(t), ..., fridge_on(t+n)]
are converted into matrix form and solved with a linear programming model.
The model's action for the next five minute interval, fridge_on(t), is rounded to 1 or 0 and used to turn the fridge on/off. While the simulation behaves well without this rounding, the discontinuous approch felt more practical and realistice given the project description.
If I continued to work on this project I would:
- move model and fridge parameters to a config file
- would lead to a cleaner api
- add more tests
- extend the forecast with a linear trend rather than a global average
- add a Mixed-Integer Linear Programming model with this repository
- add more docstrings and sphinx documentation
- add features to help compare models' performance