Solar-Forecast-ML

SFML is the first fully local AI solar forecast for Home Assistant, powered by a local Attention Transformer. No external AI — such as ChatGPT, Gemini, or Grok

What is Solar-Forecast-ML?

SFML is the first fully local AI solar forecast for Home Assistant, powered by a local Attention Transformer. No external AI — such as ChatGPT, Gemini, or Grok — required. Runs entirely on your device for complete privacy.

Overview of Solar-Forecast-ML

What it is

Solar Forecast ML (SFML) is a Home Assistant integration that forecasts solar production using a local AI — a transformer with attention — running entirely on the user's own hardware. It does not use any external AI services such as ChatGPT, Gemini, or Grok. The system learns from the user's own measurements, weather data, and system geometry to build a digital twin of the installation. All data stays on the local device; nothing is uploaded to the cloud or to the developer.

Key features

  • Runs a complete AI stack (Hubble for solar forecast, Kepler for energy decisions) locally inside Home Assistant without TensorFlow or PyTorch.
  • Produces an hourly forecast for today, tomorrow, and the day after, recalculated every morning.
  • Learns from the user's specific installation: shading, panel groups, inverter clipping, and local weather effects like snow, fog, and haze.
  • Reports measured forecast quality (accuracy, deviation, usable data days, long-term trends).
  • Pre-trained on multi-year weather and climate data, then fine-tuned on the user's system.
  • Companion modules (STATS, Energy AI, Grid) add energy workspace, explainable recommendations, and smart charging.
  • Provides six ready-made dashboard cards that follow the Home Assistant theme.

Use cases

  • Homeowners with a DC power sensor on their solar panels who want a forecast that adapts to their actual roof, not a generic model.
  • Users who require complete data privacy and do not want any solar production data sent to cloud services.
  • Systems with multiple panel groups, shading, or inverter clipping that need a model that learns these factors.
  • Home Assistant users running version 2026.3.0 or newer who are willing to understand their PV system and provide system data (kWp, orientation, tilt).

Getting started

See the repository README or the official website for installation instructions.

License and pricing

The license is not stated in the repository metadata. The project is listed as open source with no pricing.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/zara-tooroxsolar-forecast-ml.svg)](https://yunodb.com/tools/zara-tooroxsolar-forecast-ml?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/zara-tooroxsolar-forecast-ml?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/zara-tooroxsolar-forecast-ml.svg" alt="Featured on YunoDB" height="20"></a>