Welcome to F97.BE!
Python, Photons, and Predictions.
AI Energy Production Forecast
This page shows my solar energy forecast powered by two AI models trained on over 1,350 days of real-world data pulled from my Fronius inverter. Each day, two models - Random Forest and Gradient Boosting - are evaluated and compared using key performance metrics, including R² score (explaining variance) and MAE (Mean Absolute Error). My system automatically selects the best-performing model to generate the most accurate forecast possible.
By combining weather predictions with historical production patterns, the forecast estimates daily solar yield in kWh, and provides side-by-side comparisons with my GTI-based forecast and actual inverter data.
More in-depth information:
- How I use AI — and whether it actually is.
- How the AI forecast was created.
- Model Evaluation Over Time is the process of systematically measuring the performance of a predictive model over consecutive periods to detect changes in accuracy, robustness, or relevance to real-world data.
Forecast Summary
Daily Energy Production Comparison
7 Days Forecast Comparison AI and GTI
Forecast Accuracy Comparison
| Date | Actual (kWh) | AI Forecast | GTI Forecast | AI Error | GTI Error | Closer Forecast | ↑ Error Margin (%) |
|---|
Historical Forecast Accuracy
| ↓ Date | Actual Production (kWh) | AI Forecast (kWh) | % of AI Forecast Created | Model R² | MAE (kWh) | Model Used | GTI Forecast (kWh) | % of GTI Forecast Created |
|---|