The Building Thinks for Itself: How AI is Transforming Energy Management in Frickenhausen

When I arrive at the office in the morning, the building has already made its decisions: The infrared heaters on the upper floor have been running at half power since 6:30 AM because the weather forecast predicts 14 °C and sunshine from 9 AM. The battery storage was kept at 40 % overnight – enough reserve for the morning, but with room for the expected PV surplus at noon. The wallboxes in the parking lot are set to "Solar Only" because no urgent charging demand has been registered. No human configured any of this. The building thinks for itself.

Why a Commercial Building Needs a Brain

Our office building in Frickenhausen has been an all-electric site since 2012: 70 kWp photovoltaics, infrared heaters, decentralized ventilation, eight AC charging points – and soon a battery storage system with 200 kW capacity. The technology is there. But technology alone is not enough.

Because a building with PV, storage, wallboxes, and heating is a complex energy system. Every component affects the others. When the sun shines and nobody is charging, electricity flows into the grid – at prices that barely make sense anymore. When everyone charges simultaneously and clouds roll in, the building draws expensive grid power. The question is not whether you have the technology, but how intelligently it works together.

This is exactly where Artificial Intelligence comes in.

Home Assistant as the Nervous System

The backbone of our building automation is Home Assistant – an open-source platform originally developed for smart homes, but now well established in commercial settings. Through Home Assistant, all our data streams converge:

• Real-time PV generation (inverter data)
• Battery charge level and power output
• Consumption per floor and device category
• Wallbox status of all eight charging points
• Outside temperature, solar irradiance, weather forecast
• Room temperatures and ventilation status

That alone would be good monitoring. But we go one step further.

EMHASS: When Machine Learning Writes the Energy Plan

On top of Home Assistant, we run EMHASS (Energy Management for Home Assistant) – an AI module that doesn't just measure our energy consumption but optimizes it predictively. EMHASS uses three core building blocks:

1. Load Forecasting with Machine Learning
The system learns from historical consumption data when our building needs how much energy. Monday at 8 AM looks different from Friday at 2 PM. The model recognizes patterns – and plans ahead.

2. PV Forecasting with Weather Data
Through weather forecast integration, EMHASS calculates how much solar power to expect over the next 24 to 48 hours. This forms the basis for all charging and heating decisions.

3. Optimization through Linear Programming
From load forecast, PV forecast, and electricity tariffs, EMHASS calculates the optimal schedule: When does the battery charge? When do the infrared panels heat? When can the wallboxes draw full power? The result is a timetable that maximizes self-consumption and minimizes costs.

What This Means in Practice

The results after the first months: Our self-consumption rate has risen from around 55 % to over 75 %. Grid electricity costs have dropped by an estimated 30 % – not through new hardware, but through smarter use of existing equipment.

No Rocket Science: What You Need

The remarkable thing about this approach: It's not a million-euro project. Home Assistant is open source and free. So is EMHASS. The hardware – a small server or Raspberry Pi – costs under 200 euros. The real investment is time: connecting sensors, configuring automations, tuning the system to your specific building.

For a commercial building with an existing PV system and storage, the return on investment is typically achieved within 6–12 months – solely through improved self-consumption rates.

The Building as a Living Lab

Our Innovation HUB in Frickenhausen is deliberately designed as a living demonstration site. What we test here, we recommend to others – with real data, real experiences, real mistakes. AI-powered energy management is the latest chapter in this story.

And it reveals something fundamental: The energy transition for SMEs doesn't fail because of technology. PV modules, storage systems, and heat pumps are mature and affordable. What's often missing is the intelligent connection – the "brain" that turns individual components into a system.

That's exactly what AI delivers today. Not as science fiction, but as open-source software running on a mini server in the basement.

What's Next?

We're currently working on three extensions:

• Bidirectional Charging (V2G): The EVs in our parking lot should not only charge but feed energy back into the building when needed. Home Assistant and EMHASS can already model this – the matching wallbox hardware is coming to market in 2026.
• Dynamic Electricity Tariffs: With variable tariffs (e.g., Tibber, aWATTar), AI optimization becomes even more valuable. The system buys electricity when it's cheap and avoids consumption during peak times.
• Predictive Indoor Climate Control: Using CO₂ sensors and occupancy forecasts, ventilation and heating should respond not just to temperature, but to actual usage patterns.

Those who understand their building can optimize it. Those who optimize it don't just save money – they build the infrastructure for tomorrow.

Want to see how this works in practice? Visit us at the Innovation HUB Frickenhausen – we'll show you the dashboard live and discuss what's possible for your building.