Calibrated energy models for three NHS hospital buildings

Introduction

In 2026 xWatts built calibrated energy models of three NHS hospital buildings: a modern all-electric outpatient building, an older gas-heated hospital with operating theatres and wards, and a large acute hospital building served by a shared site heating plant. Each model shows where the building's energy goes and what each improvement would save before any money is spent.

‍

Approach

Each building was modelled in DesignBuilder (EnergyPlus) from floor plans, plant data and on-site sensor data, following CIBSE TM63 and ASHRAE Guideline 14. Each model was then calibrated against the building's metered electricity and gas use. All three met the Guideline 14 monthly limits, with monthly CV(RMSE) between 4.6% and 8.8% against a limit of 15%, and NMBE within ±2.5% against a limit of ±5%.

‍

What the models found

The same problem appeared in all three buildings. Air handling units ran to fixed schedules, and in some areas 24 hours a day, regardless of how many people were in the building. Demand-controlled ventilation was the largest low-cost opportunity in every model.

Modelled results from controls changes alone, across the three buildings:

• £107,000 a year in energy cost savings, at reference tariffs of 34p/kWh for electricity and 4p/kWh for gas
• 1.6 GWh a year less energy and 287 tCO₂e a year less carbon
• An estimated £19,000 to £29,000 to implement, paying back in under four months
• Total energy use cut by between 7% and 35% per building

‍

Spending capital in the right order

The models also tested capital measures, from corridor lighting sensors and smart plugs to AHU heat recovery and boiler replacement, and ranked each by cost and saving. In the largest building, triple glazing, wall insulation and airtightness works each cut gas use by only around 1%, because ventilation and hot water drive most of its gas demand. The model advised against spending an estimated £1.9m to £3.4m on them until operations are optimised.

‍

All figures are modelled from calibrated baselines and will be confirmed by measurement once changes are made. Client names are withheld.