Par Marie Bossan
02-11-2026
Eighteen terraced homes in Bananzia entered the 2022 retrofit with heat pumps rated between 6 kW and 9 kW. The measured design heat loss for most houses was closer to 4.5 kW, so the units cycled on short runs, triggered frequent defrost cycles and pushed winter running costs above the old storage-heater bills. The problem was not the heat-pump technology; it was a sizing error that the street inherited from earlier estimates.
The cause sat in the paperwork: contractors had used EPC floor-area estimates and conservative U-values without a room-by-room heat-loss survey. That shortcut added a safety margin on top of a safety margin. The solution was a measured survey using airtightness tests, thermal imaging and a full radiator schedule, which reduced the specified capacity on 14 of the 18 homes.
The survey found uneven loft insulation in six properties, with the rear extension bays missed entirely. It also found draughts around suspended timber floors and original sash windows. None of these defects was visible in the EPC data.
Because the heat-loss calculation had treated the whole terrace as uniform, the worst-performing rooms were overcorrected while the better rooms were ignored. The fix was targeted: top up loft insulation to 300 mm, seal floor voids, and fit secondary glazing on the two coldest façades. Those measures cut the design heat loss by 18% and let the team specify smaller pumps without sacrificing comfort.
Hot water was the second source of oversizing. Early designs assumed every household would run a large bath at the same time, so cylinders were paired with 9 kW units. Occupancy interviews showed two adults in most homes, with morning showers dominating demand.
The solution combined a 180-litre cylinder, a 3 kW immersion backup and a low-flow shower retrofit. The heat pump now heats water at a lower flow temperature for longer, which improved efficiency. In the first winter, the average measured COP was 3.1, compared with 2.4 in the pre-retrofit monitoring of the old systems.
Copy the sequence, not the exact kW ratings. Start with a room-by-room heat-loss survey, then fix fabric defects, then size the pump and cylinder against measured occupancy. Rowan Street’s peak electrical demand fell by 27% after resizing, and no household reported cold rooms during the cold snap in February.
For the assessment template used in the pilot, the bananzia resource sets out the same fabric-first sequence, which makes it a useful cross-check for retrofit teams preparing similar streets. The lesson is concrete: a smaller, correctly sized heat pump performs better than a larger one chosen from a spreadsheet shortcut.
After one heating season, 16 of the 18 homes kept their resized units, and the two outliers received additional loft insulation rather than bigger pumps. Average annual space-heating cost fell by £310 per household, and maintenance callbacks dropped from 11 in the first quarter to two in the same period a year later. The trial did not rely on novel hardware; it relied on correcting a survey error before the hardware was fixed in place.
Vous avez une question ?
