Underfloor heating and other slow systems: run them steadily
Slow systems punish the habits that work on radiators and a gas boiler. Whatever you asked for hours ago is what you are getting now.
Slow systems punish the habits that work on radiators and a gas boiler. Whatever you asked for hours ago is what you are getting now.
Underfloor heating, a well-insulated house, a heat pump, a heavy masonry building: all of these are slow systems, and they punish the control strategies that work on a gas boiler and a set of radiators.
The mistake is universal and understandable. People apply the habits of a fast system — big setbacks, boost when cold, schedule tightly around occupancy — and get a house that is never quite right and costs more than it should.
Here is how slow systems actually want to be run.
Two properties determine how a system behaves.
Thermal mass is how much heat the structure stores. A concrete floor slab, solid masonry walls and screed all store a great deal; a timber-framed house with a suspended floor stores little.
Lag is how long it takes a change in heat input to show up as a change in room temperature. A radiator has a lag of minutes. A screed underfloor system has a lag of hours — commonly two to six, sometimes more in a thick slab.
High mass with long lag has one enormous advantage and one enormous constraint. The advantage: temperature is remarkably stable, comfort is even, and the floor becomes a store you can charge when energy is cheap. The constraint: you cannot change your mind quickly. Whatever you asked for hours ago is what you are getting now.
The single most important principle.
A slow system should aim to hold a steady temperature continuously, adjusting output gently to match the heat the building is losing. It should not be switched on to reach a target and off again.
Why setbacks fail: if it takes four hours to raise a slab's output meaningfully, then a 6am start for a 7am wake-up does nothing, and a deep overnight setback means a cold morning followed by hours of catch-up. Worse, that catch-up usually forces a higher flow temperature, which on a heat pump directly costs efficiency — so the setback that was supposed to save money spends it.
What to do instead: pick your comfortable temperature, hold it, and allow at most a shallow overnight reduction of a degree or so. In a well-insulated house the overnight drop without any setback at all is often less than a degree anyway, which makes the whole question moot.
Since you are matching heat loss continuously, the system needs to know how much heat the building is losing — and that depends on the weather.
Weather compensation uses an outdoor sensor to set flow temperature according to outdoor conditions: cooler water on mild days, warmer on cold ones. The result is a system that runs continuously at the lowest temperature that will do the job, which is the definition of efficient operation for both a condensing boiler and a heat pump.
The compensation curve is the relationship between outdoor temperature and flow temperature, and it usually needs adjusting from the installer's default. The tuning method is simple and slow: if the house is cold in mild weather but fine when it is freezing, the curve is too shallow; if it is comfortable in mild weather and cold in a hard frost, it is too steep. Change one setting, wait two or three days, then judge. Changing it daily makes tuning impossible, because the building has not caught up with the last change.
Room influence — letting an indoor sensor nudge the curve — is a useful refinement, and it should nudge rather than override.
Room-by-room zoning is much less useful on a slow system, and can be counterproductive.
Underfloor heating in one room warms the slab, which conducts sideways and radiates upward for hours. Turning a zone “off” does not produce a cold room quickly, and turning it on does not produce a warm one quickly, so occupancy-based control has little to work with.
Closing zones also reduces the emitting surface available, which pushes flow temperature up — the same trap as over-zoning a heat pump.
What does work: small persistent differences between rooms, set once and left. A bedroom a degree or two cooler, a bathroom a degree warmer. Set it, let the building settle over a few days, and stop adjusting.
Many houses have underfloor downstairs and radiators upstairs, and the two want different water temperatures — underfloor typically 35 to 45 °C, radiators traditionally much higher.
The usual arrangement is a mixing valve or a blending manifold that takes hot water from the boiler and blends it down for the floor circuit, with the radiators fed directly. It works and it means the boiler still runs hot, giving up the efficiency benefit of low-temperature operation.
The better arrangement, where possible, is to bring the whole system down by upsizing the radiators so they work at the lower temperature the floor wants. That is the approach used in heat pump design and it produces a system that is efficient throughout.
Control them separately in either case: the radiators can respond to a schedule, the floor should hold steady.
Here is where a slow system becomes an advantage rather than a limitation.
A heavy floor is a heat store. On a time-varying tariff you can put heat into it during cheap hours and coast through expensive ones, and the occupants feel nothing because the slab smooths it out entirely. Combine that with solar generation in spring and autumn and you have a genuinely useful flexible load.
Practically: raise the target by a degree during the cheap window and let it drift back afterwards. Keep the swing small, because a large overshoot is uncomfortable and inefficient, and the point is a gentle nudge rather than a charge cycle.
Note the interaction with weather compensation: a fixed cheap-hour boost is crude in mild weather. Scaling the boost with outdoor temperature, or skipping it when the forecast is mild, avoids overheating the house on a warm February night.
An underfloor system's output depends on how easily heat gets through what is on top of it, expressed as thermal resistance.
Stone and tile pass heat well and are ideal. Engineered wood is usually acceptable within limits set by the manufacturer. Thick carpet with a heavy underlay can insulate the floor to the point where the system cannot deliver its designed output, and the response — raising flow temperature — costs efficiency and may still not be enough.
If a room with underfloor heating never gets warm enough, check the floor covering before suspecting the system. It is a more common cause than most people expect, and changing a rug is cheaper than changing a heat pump.
The last point is the one that matters most and the hardest to follow. Almost every complaint about underfloor heating and heat pumps traces back to someone adjusting a slow system on a fast timescale, which guarantees it never settles.