The problem
A tiled floor on 10 cm of concrete is a slow, heavy heat battery. Most thermostats treat it like a lamp with a timer: warm at some hour, cooler at another, and no idea what electricity costs at that moment.
The spot price does not follow a timer. On one day in early October it went from about 0.19 DKK/kWh around noon to 1.33 in the morning rush. A floor that can hold heat for hours does not need to be heated at the worst possible time.
At 1.200 W and roughly 70 kWh a month, this will not change a household budget. That is fine. It is a small, contained system that is safe to experiment on, and the interesting part is whether the logic holds up. It is also a small example of the idea behind Software as I Want It: instead of waiting for an app to offer price control, describe what the floor should do and build it.
The decisions
Start from the data, not from a round number. My own starting point was that 2 DKK/kWh counts as expensive. The price sensor's highest value that day was 1.86, and the next day peaked at 1.33. A threshold of 2 would never have fired, and the automation would have done nothing without telling me. For the first test, "cheap" is 1.00 DKK/kWh or less and "expensive" is 1.25 or more.
Three steps instead of on and off. Cheap power sets the floor target to 27 °C, mid-price to 22 °C and expensive power to 15 °C. The concrete does the real work in between.
Hourly average instead of quarter-hour prices. Reacting to every 15-minute price would make the thermostat nervous. The average of the current hour keeps it calm.
A safety net for moisture. An aggressive cool-down can leave a bathroom floor cold and damp. If the floor drops below 17 °C, the automation overrides the price and brings it back up to 19 °C. It protects the building before it protects the budget.
Fail soft. No price data, or the price control switched off, and the thermostat falls back to its own schedule. If the thermostat is offline, the automation does nothing. A smart home that makes silly decisions when something breaks is worse than a dumb one.
What was built
A small home-automation setup with:
- Home Assistant running as a virtual machine on Proxmox, on a small home server
- The DEVIreg Smart thermostat paired over MQTT, using Mosquitto and a community-built add-on
- Live Nord Pool spot prices for the DK2 price area, installed through HACS
- One automation that turns the current price into a heat step and a floor target, checked every 15 minutes
- A master switch that hands the thermostat back to its own schedule
- A moisture safety net and fallbacks for missing data
Where it stands
The first test is deliberately aggressive, and it is meant to be adjusted from what the data shows. The key number is how fast the floor loses temperature at the low setpoint, because that decides how long it can coast. Kilowatt-hours per day will show whether any of it matters.
One thing is already clear: the Nord Pool spot price is only part of what electricity costs. It leaves out the grid tariff and taxes, and the grid tariff usually rises in the evening peak. The next test will use the total price instead.
Concrete is slow. That is the whole trick.
If you like small, very specific problems, Enebakken is the opposite end of the smart-home scale: a Swedish house with no electricity at all.