S2Cool: Digital-Twin Modeling for Refrigerant-Free Evaporative Cooling

A £2.8M UKRI-Funded Ayrton Challenge Consortium Research Project

Author(s)

Muhammad Hammad Irfan

Academic / Industry Supervision

Dr. Moazam Fraz (NUST SEECS), Dr. Naseer Bajwa (NUST SEECS)

Institution & Collaborators

National University of Sciences and Technology (NUST) • Northumbria University (UK) — consortium lead • 12+ additional UK and Pakistani partner institutions

UKRI Ayrton Challenge Consortium

Timeline & Funding

2026 – Present

UK Research and Innovation (UKRI), Ayrton Challenge Programme — £2.8M consortium grant

Interactive Demo & Screenshots

Consortium Funding

£2.8M

UKRI Ayrton Challenge Programme

Empirical Dataset

~66k Points

Sensor readings from 11 instrumented prototype test runs

Model Accuracy (current)

R² ≥ 0.93

Supply-temperature and thermal-effectiveness prediction — ongoing work, not finalized

Prediction Error (current)

RMSE < 0.5

Across tested ambient temperature and humidity conditions

Abstract & Executive Summary

Space cooling in extreme-heat climates is dominated by vapor-compression air conditioning, which carries a heavy electrical load and relies on high-GWP refrigerants. S2Cool is a £2.8M UKRI Ayrton Challenge consortium, led by Northumbria University with NUST SEECS among 12+ UK and Pakistani partner institutions, developing a refrigerant-free evaporative cooling system targeting up to 65% lower cooling energy demand. As part of the NUST SEECS team, this ongoing work contributes to the project's digital twin: a dual-track behavioral model combining first-principles heat and mass-transfer equations with data-driven ML surrogate models, trained on roughly 66,000 empirical observations from 11 instrumented test runs on the physical prototype. Current Ridge regression surrogate models reach an R² of at least 0.93 and an RMSE under 0.5 for supply-temperature and thermal-effectiveness prediction — figures that reflect current, ongoing progress rather than a finalized result.

#Digital Twin#Applied Machine Learning#Clean Energy#Sustainable Cooling#UKRI Research

1. Research Motivation

Reducing Cooling Energy Demand in Extreme-Heat Climates

Cooling is one of the fastest-growing sources of energy demand in buildings, and extreme summer temperatures in regions like South Asia (exceeding 52°C) place heavy strain on electrical grids reliant on conventional refrigerant-based air conditioning. Refrigerant-free evaporative cooling offers a path to substantially lower energy demand without synthetic refrigerants, but designing and validating such systems across varied climates requires extensive performance modeling — motivating a digital-twin approach rather than physical prototyping alone.

Key Insights & Methodology

Conventional air conditioning drives significant peak electrical demand in extreme-heat regions.
Refrigerant-free evaporative cooling avoids high-GWP synthetic refrigerants.
Digital-twin modeling reduces the cost and time of physical prototype iteration.

2. Dual-Track Digital Twin Architecture

Combining First-Principles Modeling with Data-Driven Surrogates

Purely analytical thermodynamic models can miss real-world hardware effects such as uneven wetting of the evaporative matrix or airflow non-uniformity. A purely data-driven model, on the other hand, risks fitting patterns that don't respect real physical constraints. The project combines both: - **First-principles track**: heat and mass-transfer equations governing the evaporative cooling process. - **ML surrogate track**: Ridge regression models trained on real sensor telemetry to capture behavior the analytical model alone doesn't fully account for.

Wet-Bulb Effectiveness

A standard psychrometric measure of an evaporative cooler's performance: ε_wb = (T_dry-bulb,in − T_dry-bulb,out) / (T_dry-bulb,in − T_wet-bulb,in) Surrogate models predict outlet dry-bulb temperature, from which effectiveness is derived and checked against the physical bound 0 ≤ ε_wb ≤ 1.

3. Physical Prototype Data

~66,000 Empirical Observations from 11 Test Runs

Data was collected from physical, sensor-instrumented test runs on the prototype cooling bench in London, across varying ambient temperature and humidity conditions, producing roughly 6,000 data points per component run and approximately 66,000 observations in total.

Key Insights & Methodology

Real sensor data, not simulated data, underlies the surrogate models.
Testing spans varying ambient temperature and humidity conditions.
Exact instrument specifications are not listed here — confirm real sensor/precision details from the lab setup before publishing specific hardware specs.

4. Current Model Performance

Ongoing Work — Figures Reflect Current Progress

Ridge regression surrogate models are trained per subsystem (humidifier, wet channel, dry channel) against real sensor data. Current results reach an R² of at least 0.93 and an RMSE under 0.5 for supply-temperature and thermal-effectiveness prediction. As this is an active, ongoing research project, these figures represent current progress rather than a final, published result, and will be updated as work continues.

Key Insights & Methodology

Results are reported at the aggregate level (R² ≥ 0.93, RMSE < 0.5) rather than with false per-metric precision, since the work is still in progress.
Multiple modeling approaches were evaluated per subsystem, with Ridge regression performing best so far.

Cite This Technical Report / Research

@misc{irfan2026s2cool,
  title={S2Cool: Digital-Twin Modeling for Refrigerant-Free Evaporative Cooling (Ongoing)},
  author={Irfan, Muhammad Hammad},
  howpublished={NUST SEECS, UKRI Ayrton Challenge Consortium},
  year={2026}
}

View Visual Overview & Media Explore All Projects