The prediction service predicts the daily average energy price in Stockholm/SE3 for the upcoming 7 days. The service is of value to people living in the SE3 elområde and who want to know the electricity prices for the upcoming week.
XGBoostRegressor, which we performed hyperparameter tuning on.
The entsoe API (Transparency Platform RESTful API) is used to fetch data about energy prices, load, and aggregated filling rate of water reservoirs and hydro storage plants. For this we used the entsoe-py client.
Visual crossing weather API is used to featch weater data, including: temperature, wind gust, wind speed, wind direction, and cloud cover
elbruk.se dagspris is scraped daily through beautifulsoup and a modal job
EUR and SEK currency rates are collected from a python library that get daily rates from the European Central Bank.
The date features need no external data source.
- The prediction service solves a useful problem
- We've added a detailed monitoring page for the prediction service
- We've used data validation using Pandera to validate the data before adding it to the feature groups
- Reused code is in the utils module to follow best MLOps practices. Data validation schemas are read from file to avoid code duplication and training/serving drift.