California ISO Model for load forecasting (CAISO): It predicts for the next day by taking hourly averages of the three days with highest average consumption value among a pool of ten previous days, excluding weekends, holidays, and past DR event days
Tho model is packed as a MLflow Project. When run, it fetches the newest data, trains a CAISO model and saves it locally as an MLflow Model. Here's a starting point for running and deploying a CAISO model locally:
git clone https://github.com/NielsOerbaek/caiso-mlflow
cd caiso-mlflow
mlflow run .
mlflow models serve -m model
You can now query the model using curl, for example:
curl http://127.0.0.1:5000/invocations -H 'Content-Type: application/json' -d '{
"columns": ["Time"],
"data": [["2020-11-14T20:00:00"]]
}'