Electricity-Price-Forecast

Electricity Price Forecast

AI-powered predictions for the Stockholm Region (SE3).
Optimized for smart energy usage.

Next-day forecast

Tomorrow’s electricity price predictions (early access)

Official day-ahead prices are typically published around 13:00. This dashboard provides an earlier forecast so you can plan flexible consumption (EV charging, laundry, dishwasher) before the official release. After 13:00, use official prices as the source of truth.

Forecast date:
Updated (UTC):

Cheapest hours

Good time for EV charging, laundry, dishwasher.

Most expensive hours

Try to avoid flexible consumption in these windows.

Cheapest 4-hour window (charging)

The lowest average price across 4 consecutive hours tomorrow.

10

Estimated savings:

The slider only affects the savings estimate (it scales with how many kWh you can shift).

Tomorrow: hourly predictions

Tomorrow hourly predictions

How to use: Use the green/amber/red bars as a quick guide for flexible consumption. Combine this with the “Cheapest hours” and “Cheapest 4-hour window” cards above for an actionable plan.

Model reliability (recent history)

Price Trend

How to read: Black is the actual price, orange dashed is the model prediction over recent days. This helps validate that the model tracks price dynamics before using it for planning.

What drives the prediction?

Feature Importance

Understanding the graph: Higher bars mean the model relies more on that feature. Lag features (e.g. price_lag_24) capture that prices often repeat daily patterns.

About the Project

This project forecasts next-day hourly electricity prices for SE3 before the official day-ahead prices are published (~13:00). The goal is to support morning planning for flexible consumption (EV charging, laundry, dishwasher).

Data sources (APIs)

  • Electricity prices: day-ahead hourly prices for Sweden price areas (SE1–SE4) from elprisetjustnu.se (via the proxy API at api.elpris.eu when available).
  • Weather: hourly forecast and historical weather from Open‑Meteo.

Note: after 13:00, official prices should be used as the source of truth.

Pipelines

Daily ingestion updates the feature store, daily inference generates predictions and dashboard assets, and monthly training retrains the model.

Technology

Python, XGBoost, Hopsworks Feature Store & Model Registry, GitHub Actions, GitHub Pages.

View Code on GitHub
© 2025 Scalable Machine Learning Project.