April 15, 2026

Noviterm AI Achieves up to 40% Cost Reduction in Cross-Validated Benchmark

We are excited to share the results of our latest benchmark validation. Before activating Noviterm AI on any real building, we build a precise Digital Twin — a physics-informed AI model that learns the building’s exact thermal behaviour, wall structure, solar exposure, and HVAC topology.

To guarantee performance, we cross-validate this Digital Twin against two of the world’s most trusted engineering simulation engines:

🔬 EnergyPlus — the U.S. Department of Energy’s flagship building simulation standard, used worldwide for energy compliance calculations.
🔬 OpenModelica (FMU/FMI 2.0) — an open-standard dynamic systems modelling platform used in industrial control engineering.

Only when both engines confirm the model’s accuracy do we activate live optimization.

The numbers below show what this process delivers on the ASHRAE Large Office reference building — the global standard for commercial energy research. The Noviterm AI controller (Neural ODE + MPC) was tested against the industry-standard ASHRAE Guideline 36 PID controller.

Winter Results

−20%
⚡ Electrical Energy
−26.9%
💰 Cost Savings
−8.5%
♨️ Thermal Energy
<0.4°C
🌡️ Comfort MAE

3.5 days · 15 zones · Nordpool-extreme tariff vs ASHRAE G36

Summer Results

−25.1%
⚡ Electrical Energy
−40.0%
💰 Cost Savings
−21.3%
❄️ Thermal Energy
<0.7°C
🌡️ Comfort MAE

3.5 days · 15 zones · Nordpool-extreme tariff vs ASHRAE G36

Methodology

All simulations ran with Nordpool-extreme electricity tariff profiles to validate dynamic pricing optimization capabilities. The benchmark covers both heating-dominant (winter) and cooling-dominant (summer) scenarios.

Results validated in industrial-grade simulation (EnergyPlus + OpenModelica) on ASHRAE Large Office standard. Individual results will vary by building type and HVAC system.

← Back to News