Presets
Playback
Climate and house
Heat pump and thermostat
Experiment
Camera
Layers
Quality
Model check
The science
The model: three lumped heat stores
The house is reduced to three temperatures: the living-room air with its furnishings (T1), the bedroom (T2) and the building structure (Tm, walls and slab). Each store obeys C dT/dt = sum of heat flows. For the living room:
C1 dT1/dt = H1 (To - T1) + Um1 (Tm - T1) + U12 (T2 - T1) + share * Qhp + Qsolar + Qinternal. H is the loss coefficient to outdoors in W/K: walls A U, windows, roof or ground floor, plus ventilation rho cp V n / 3600 with n air changes per hour. Opening the window adds 6 air changes per hour, which roughly triples the living-room loss coefficient. The structure store has a large capacity (about 8 MJ/K), so the house cools slowly: the overall time constant is tau = sum(C) / sum(H), shown live (typically one day).
Heat pump
An air-to-water heat pump moves heat with an electrical input P = Q / COP. A real machine reaches a fraction eta of the Carnot limit, COP = eta Th / (Th - Tc) with Th the flow temperature and Tc the outdoor temperature in kelvin; here eta = 0.42, so COP is about 4.0 at 7 C outdoors with 40 C flow, and about 2.5 at -10 C. That is why low-temperature radiators (a lower flow temperature) cut the electricity bill. The heat output falls with outdoor temperature (2 percent per kelvin below 7 C in this model); the "Undersized pump" preset shows what happens when it can no longer match the heat loss. Below 2 C the COP is reduced by 8 percent to account for defrost cycles.
Thermostat and strategies
The controller is a bang-bang thermostat with hysteresis: it switches on below Tset - h and off above Tset + h, with a 3 minute compressor lock-out. Fixed keeps one setpoint all day. Schedule lowers it at night (22:30-06:30). Smart also drops to the eco setpoint when nobody is home, pre-heats one hour before people return or wake, and switches the heat pump off two minutes after a window is opened. The "Compare strategies" button simulates four days of each strategy with the same weather and reports the last day, so the saving is measured, not asserted.
Simplifications, stated honestly
- Three nodes only: no room-by-room detail, no stratification, radiators at a fixed flow temperature and a fixed 65/35 heat split between floors.
- Solar gain is a simple half-sine of irradiance on the south glazing (peak 450 W/m2 times the cloud factor); half the gain goes to the structure.
- Ground coupling is a constant factor; no thermal bridges, moisture or hot water.
- Tariff 0.25 per kWh and grid emissions 0.2 kg/kWh are illustrative numbers, not a quote. U-values, heat capacities and COP efficiency are typical values for a modern semi-insulated house; check them against your own building before drawing conclusions.
The independent checks on this page
- Cooling: with the heat pump off, the real integrator must reproduce
T = To + (T0 - To) e^(-t/tau); after one tau it must be 36.8 percent of the starting difference. - Steady state: with a constant 2 kW supplied the temperature must settle at
To + Q/H, with H summed by hand from areas and U-values. - COP: the Carnot-fraction formula against the hand value for 7 C outdoors and 40 C flow.
- Energy balance: stored energy change equals cumulative heat gains minus losses, which also proves the heat flowing between rooms cancels.
Assets: the house, furniture, heat pump and radiators are modelled in Blender, with baked PBR textures, and driven by Three.js. This is an educational model, not building-regulation software.