Floor heating, radiators and weather compensation

A heat pump is only as efficient as the water temperature it has to deliver. That supply temperature (also called flow temperature) is the temperature of the water the heat pump sends into your heating circuit. Two things set it: the emitters in the rooms, meaning whatever hands heat from the water to the air, such as floor pipes or radiators, and the control that adjusts it through the year. Both change the result.

Why a cooler supply temperature helps

A heat pump does not make heat. It lifts heat from the outdoor air up to the water in your heating circuit, a bit like pumping water uphill. The hotter the water, the longer the climb and the more electricity each unit of heat costs. The measure is COP, the coefficient of performance: kWh of heat per kWh of electricity. At a COP of 4, one kWh of electricity gives four kWh of heat, where a plain electric heater gives one.

In the model, the heat pump’s COP at each supply temperature comes from the manufacturer’s published tables for water at 35, 45, 55 and 65 °C, not from a correction formula. For one mid-range unit at an outdoor temperature of 7 °C, the published COP is 5.2 with water at 35 °C, 3.7 at 45 °C and 2.7 at 55 °C. The model reads the table at the supply temperature the circuit delivers in each hour, so a 50 °C radiator circuit sits between the 45 and 55 °C columns.

Emitters decide the supply temperature

An emitter has to move heat from the water into the room. The larger its surface, the cooler the water can be. You choose the emitter type in the wizard, and the design supply temperature (what the system is designed to run at on the coldest days) follows from it.

Underfloor heating uses the whole floor as a radiator. It heats well with water in the low 30s of °C, and the model uses 35 °C. Low-temperature radiators are sized for a cooler regime than old boiler radiators, and the model uses 50 °C. High-temperature radiators are what a boiler house usually has, and the model uses 65 °C.

Radiators are rated by the EN 442 reference condition of 75/65/20 °C (water in, water out, room). At heat pump temperatures the gap between the radiator’s average water temperature and the room is far smaller, so the same radiator gives noticeably less heat. That is why a house that moves from a boiler to a heat pump often needs bigger radiators or a higher supply temperature, and a higher supply temperature costs COP.

Weather compensation

On a mild day the house needs far less heat than on the coldest day, so the water does not need to be as hot. Weather compensation is the control that lowers the supply temperature automatically as the outdoor temperature rises. The rule it follows, outdoor temperature in and water temperature out, is the heating curve (in German, Heizkurve).

Two things improve. The heat pump works against a smaller temperature lift, so COP rises. And a heavy concrete floor, which holds a lot of heat and releases it slowly, is not fed more than the room needs, so the room is less likely to overshoot the setpoint (the temperature you asked the thermostat to hold).

In the model, an underfloor loop follows a heating curve. The supply temperature is the design value (35 °C) at an outdoor temperature of -10 °C or colder, falls in a straight line to 25 °C at 15 °C outdoors, and stays at 25 °C above that. A radiator loop follows a heating curve too. It starts at its design supply temperature (50 °C or 65 °C) on the coldest days and falls as the weather warms, down to a floor of 35 °C. The curve comes from how a radiator gives off heat, which grows a little faster than the gap between the water and the room (the EN 442 law, with an exponent of 1.3), and from the room temperature you set, so a warmer room needs warmer water. It is derived, not fitted. The heat pump tables stop at 35 °C, so the model keeps the 35 °C figures below that: a radiator loop earns a better COP on mild days, and a floor loop, whose design supply is already 35 °C, earns no credit for its cooler water. You cannot change the slope (how fast the water cools as the weather warms) or offset (a shift of the whole line) of either curve.

How the model sizes emitters from the house

An emitter sized for more heat than the heat pump can supply asks for heat that does not exist, and the backup heater (an electric element that tops up the heat pump) fills the gap at a COP of 1. So the model sizes the emitters with the same design-load method that sizes the heat pump.

Each heated storey gets a design heat load: the heat it loses on the design day, a very cold day that heating is sized for. That is also what its emitters must deliver. The model sums area times U-value (how well a surface keeps heat in; lower is better) for walls, roof, floor and windows, adds an allowance for thermal bridges (weak spots such as corners, where heat escapes faster) and the heat carried out by ventilation, then multiplies by the gap between 20 °C indoors and the design outdoor temperature, which is -12 °C by default.

The bridge allowance is 0.05 W/m²K for houses built from 2015 and 0.10 W/m²K for older ones, following DIN/TS 12831-1, the German annex to EN 12831-1, the European heating-load method. That standard’s optional reheat allowance, an extra margin for recovering quickly after a night setback (a lower temperature overnight), is left out. It can oversize a well-insulated house severely, and an oversized heat pump switches on and off more often and costs more to run.

Each storey’s floor is given a capacity (how much heat it can deliver) equal to its design load. Each storey has one aggregate radiator, because EnergyPlus, the building simulation program the model runs on, treats several radiators in a room like one of the same total output. It is sized to deliver the design load at its design supply temperature. The wizard then proposes a heat pump size from the house.

Cast iron or steel panel radiators

The wizard asks what your radiators are made of, because the metal and the water inside them hold heat and keep giving it after the valve shuts. A cast iron column radiator, the heavy kind fitted mostly between 1910 and 1970, holds about 66 joules per kelvin for each watt of rated output. A steel panel, common since the 1960s, holds about 24. Those are medians over manufacturer datasheets, 50 cast iron models and the panels of six makers. The time constant, the time a radiator takes to lose about two thirds of its excess temperature over the room, is about 60 minutes for cast iron and about 20 for steel panel, at a fifth of rated output.

The type changes timing. Radiator and heat pump sizes stay as they were. After the morning setback the model brings a room back to its setpoint in about 15 minutes with steel panels and about 45 with cast iron. Over a year the energy at the meter rises by 0.5 to 3.5 % against the same house with radiators that store nothing, and with cast iron the boiler’s or heat pump’s hourly peak is about a third higher, because the metal has to be warmed again.

If you are unsure which to pick: extra-wide ornate cast iron holds about twice as much as a standard column, and steel column radiators sit close to cast iron, so choose cast iron if yours are steel columns. The model uses one type for the whole house. It also caps each radiator at its rated output, where a real cold radiator briefly gives more, so the cold mornings it shows with cast iron are an upper bound. A real house can see far fewer of them.

Limits of the model

Emitters are sized exactly to the design load. You cannot enter the radiators or floor circuits you already have.
The model has one emitter per storey, not one per room.
The ventilation part of the design load assumes heat recovery (the ventilation unit passes warmth from outgoing air to incoming air) for houses built from 2015 and none for older houses, whatever ventilation system you choose.
The heating curves are fixed in shape. The radiator curve follows the room temperature you set, and the model draws it as a straight line that never falls below the derived curve, so it can supply the water a little warmer than needed.
The heat pump tables run from 35 to 65 °C. Below 35 °C the model keeps the 35 °C figures, and above 65 °C it refuses the setup. A radiator circuit at 50 °C therefore gets a lower COP than a 35 °C floor circuit, as it does in a real heat pump. Check the model’s COP against a datasheet at your supply temperature.