Welcome to F97.BE!

Is It Really AI?

The other day, someone asked me whether phrases like “AI Energy Production Forecast” on my site were truly powered by artificial intelligence or if “AI” was just a buzzword. It is a fair question. The short answer is yes, it really is AI, but in a very practical and grounded sense.

What Powers My Forecasts

At the heart of my system is machine learning, a branch of artificial intelligence that allows computers to learn from experience. Instead of following fixed “if…then” rules, the system studies historical data — years of solar production from my Fronius inverter, combined with detailed weather records — and learns how different conditions shape daily output.

Every 15 minutes, a custom Python script polls my inverter for production data. Another script fetches weather observations and forecasts from the Open-Meteo API — temperature, cloud cover, wind speed, solar radiation, and precipitation. Over time, these datasets merge into a detailed record of how my panels respond to changing skies.

Using supervised learning, I feed the models both the inputs (weather conditions) and the outputs (actual energy produced). Algorithms such as Random Forest and Gradient Boosting then detect patterns, some obvious and others too complex for a human to define. I evaluate each model with metrics like R² and MAE, choosing the one that performs best for current conditions.

Once trained, the model takes tomorrow’s weather forecast and predicts production, a kind of digital intuition based not on guesswork but on learned cause-and-effect.

Why I Call It AI

This is narrow AI, intelligence designed for a specific task: predicting solar output with increasing accuracy and adaptability. It does not think or feel, but it interprets inputs, learns from them, and makes informed predictions. That is intelligence in the functional sense, built not by hardcoding rules but by letting the system build its own understanding from real-world data.

There is something quietly profound about giving a few Python scripts the ability to anticipate tomorrow and to improve with each new sunrise.

Why It Matters

For me, forecasting is not just about numbers, it is about control. In Belgium, solar owners pay a prosumer tariff that treats the grid as a virtual battery, storing excess power during the day and delivering it back when needed. By 2030, my goal is to make my system — panels, smart meters, electric vehicles, and storage — fully self-managing so I no longer rely on that model.

A reliable forecast is essential for that future. It turns energy production from a passive process into something I can plan around, optimise, and align with my values of sustainability and autonomy.

Looking Ahead

When I say my forecasts are powered by AI, I mean it in the most direct sense. It is applied artificial intelligence, a system that listens, learns, and adapts, helping my home run smarter every single day.

If you are thinking about your own solar setup, your grid, or your place in the energy transition, I hope this shows what is possible now and what even smarter systems will achieve in the years ahead.

Erik Schmidt, August 2025.