SolarForecast ML v3.0.0: Adaptive and Self-Learning for Pinpoint Solar Predictions

Share your Projects! Custom Integrations
, , ,
1

:tada: Solar Forecast ML v3.0.0 - Final Release

:dart: Core Features

Intelligent Forecasts

  • :white_check_mark: Forecast for Today & Tomorrow
  • :white_check_mark: Time-based Correction - “Today” reduces during the day
  • :white_check_mark: Intelligent Night Detection - No 0 kWh forecasts
  • :white_check_mark: Hourly Forecasts (optional) for short-term predictions

Machine Learning

  • :brain: Automatic daily learning (23:00)
  • :bar_chart: Day Profile Learning collects hourly data
  • :dart: Accuracy Tracking shows model performance
  • :arrows_counterclockwise: Self-Calibration adapts to your system
  • :chart_with_upwards_trend: 30-Day Profile learns typical yield curve

Config Flow UI

  • :sparkles: No YAML configuration needed!
  • :earth_africa: Multilingual (German & English)
  • :memo: Helpful Descriptions for each sensor
  • :wrench: Reconfigure - Sensors can be changed later
  • :control_knobs: Options Flow - Notifications customizable

Weather Integration

  • :de: DWD (German Weather Service) - Optimized and preferred
  • :earth_africa: Met.no - Norwegian Weather Service
  • :cloud: OpenWeatherMap - Alternative
  • :arrows_counterclockwise: Generic - Works with almost all weather integrations
  • :rocket: Auto-Detection - Automatically detects best forecast method

:bar_chart: Sensors

Sensor What it does
Forecast Today Daily forecast (time-corrected)
Forecast Tomorrow Prediction for next day
Accuracy Model performance in %
Status Integration status & diagnostics
Next Hour Short-term forecast (optional)
Manual Button Forecast on button press

:wrench: Technical Features

Stability

  • :zap: Async I/O - No blocking operations
  • :bug: Race-Condition Fix - Weather loads before Solar-ML (after_dependencies)
  • :lock: Error Handling - Robust error handling
  • :memo: Detailed Logging - Easy debugging

Code Quality

  • :package: Modular Structure - Cleanly organized
  • :dart: Entity Registry - Correct HA integration
  • :floppy_disk: Persistent Data - JSON-based storage
  • :2nd_place_medal: Quality Scale: Silver - High code quality

Performance

  • :rocket: Optimized Updates - Only when necessary
  • :dash: Fast Start - Async Loading
  • :bar_chart: Efficient Data Storage

:gift: Advanced Features

Optional Sensors (improve accuracy):

  • :sunny: Lux Sensor - Brightness in Lux or W/m²
  • :thermometer: Temperature Sensor - Outdoor temperature in °C
  • :dash: Wind Sensor - Wind speed
  • :high_brightness: UV Index - UV radiation (0-11)
  • :bar_chart: Forecast.Solar - Comparison with external forecast
  • :zap: Current Power - For day profile learning
  • :electric_plug: Inverter Monitoring - Detects offline status

Notifications (Optional):

  • :sunrise: Daily Forecast (6:00) - Today & Tomorrow
  • :crescent_moon: Learning Result (23:00) - Shows learning progress
  • :warning: Inverter Warning - Reports offline
  • :white_check_mark: Start Notification - Confirms setup

:chart_with_upwards_trend: Recap (THANKS TO THE BETA TESTERS)

Feature v2.x v3.0.0
Config :x: YAML :white_check_mark: UI (Config Flow)
Today Forecast at 6 AM :x: 0 kWh :white_check_mark: Correct prediction
Time Correction :x: None :white_check_mark: Dynamic
Day Profile :x: None :white_check_mark: 30-Day Learning
Hourly Forecast :x: Basic :white_check_mark: ML-based
Weather Support :warning: Limited :white_check_mark: Universal
Multilingual :x: English only :white_check_mark: DE + EN
Help Texts :x: None :white_check_mark: Detailed
HACS :warning: Manual :white_check_mark: Fully compatible
Error Handling :warning: Basic :white_check_mark: Robust

:package: Installation

Via HACS (Recommended):

  1. HACS → Integrations → ⋮ → Custom Repositories
  2. Add: Zara-Toorox/ha-solar-forecast-ml
  3. Category: Integration
  4. Install “Solar Forecast ML”
  5. Restart Home Assistant
  6. Add integration (UI)

Manual:

  1. Download solar_forecast_ml.zip
  2. Extract to /config/custom_components/
  3. Restart Home Assistant
  4. Add integration (UI)

:books: Documentation

  • :open_book: README: Full instructions
  • :memo: CHANGELOG: All changes
  • :bug: Issues: Report bug
  • :speech_balloon: Discussions: Questions & Ideas

:pray: Thanks

Special thanks to:

  • Home Assistant Community
  • HACS Team
  • All Beta Testers

:dart: Next Steps

After installation:

  1. :alarm_clock: Wait 24h - First data collection
  2. :bar_chart: After 7 days - First accuracy values
  3. :dart: After 30 days - Optimal performance

:star: If you like v3.0.0, give the project a star!

:bug: Problems? → Create an issue

Good luck with v3.0.0! :sun_with_face:

6 Likes
2

Instaelled without problems. Lets see how accurate it will be.

3

seems to be an issue with the language
my system language is English but sensor output data are in German

1000085191
10000851911076×2072 144 KB

1 Like
4

There is nothing quite like reading the documentation.

Entities

The integration creates the following sensors and buttons. (Note: Default entity names are in German as defined in the code; you can rename them in Home Assistant.)

5

I understand, i said the sensor output is also in German that part I can’t rename.

if that is intended to be also in German then it wlil be difficult for none German audience of you want to attract more international users

6

Ah yes. That requires template sensors to do the translation ad-hoc.
Perhaps the repo owner, or some kind soul, will add the required translation JSON.

7

also the notification are in German

Solar-Prognose
Heute: 9.6 kWh, Morgen: 8.3 kWh

:white_check_mark: SolarForecastML gestartet
Basiskapazität: 29.75 kWh

:bulb: Lern-Ergebnis 2025-10-25
Prognose: 9.63, Tatsächlich: 6.60, Abweichung: -45.9%

:brain: Modell hat für 2025-10-25 gelernt
Die Prognoseabweichung betrug -3.03 kWh. Die Gewichte wurden angepasst.

1 Like