FORGE

Writing people back to architecture and energy systems for various futures.

A design-science lab for human-building-climate systems at The University of Hong Kong.

Future, Occupant, Risk, and Generative Environments

What we're doing

Current focus

The current agenda is concentrated around standards revision, future weather uncertainty, thermal comfort prediction, and AI-supported design pedagogy.

  • ASHRAE 1959-TRP
  • Human Heat Standards
  • Climate Stress-Testing
  • Ordinal Thermal Comfort
  • Risk-Aware Control
  • Socratic Oracle

Why FORGE?

FORGE

The lab is built on a simple position: architecture and energy systems have drifted too far from the people they claim to serve. FORGE writes people back into simulation, control, climate risk, and design reasoning.

Who’s here

Current team

FORGE is a compact research group spanning a postdoctoral researcher, PhD students, research assistants on site at HKU and remote locations, and undergraduate RAs. The team is currently working across ASHRAE 1959-TRP on updating real human heat assumptions, climate stress-testing and stochastic future weather modeling, ordinal thermal comfort, and Socratic Oracle as a design-pedagogy platform.

Open Team

What we work on

Method signals

FORGE moves between building simulation, comfort science, climate futures, and design-facing AI through a small set of recurring technical workflows.

  • EnergyPlus
  • Climate Forcing
  • Monte Carlo
  • Physics-Informed Neural Networks
  • Mean Radiant Temperature
  • ERA5 / CMIP6
  • Hidden Markov Models
  • State-Based Inference
  • Ordinal Learning
  • Co-Simulation
  • Stochastic Weather
  • Sensitivity Analysis

Recent output

Selected publications

Sustainable Cities and Society · 2026

Automated Urban Energy Assessment: From Thermal Flyover to AI-Driven Retrofit Prioritization for Sustainable Cities.

Uses Thermal-SAM to segment mid-wave infrared flyovers into building patches, pairs ring and context thermal features with EPC-derived EUI benchmarks, and shows thermal data is most useful for triaging anomalous high-EUI or EPC-inconsistent buildings for follow-up audits rather than routine direct EUI regression. DOI

Hongshan Guo, Sebastiano Anselmo, Maria Ferrara, Shuai Niu, Binlin Chi, Xuchen Wang

energyclimate

Building and Environment · 2026

From Seven Points to Probabilities: Ordinal Learning for Risk-Aware Thermal Comfort Prediction.

Reframes thermal sensation prediction as an ordinal learning problem with calibrated probabilities for risk-aware comfort decisions. The model yields confidence-aware outputs that are more useful for control and design decisions under uncertainty than a flat seven-point label alone. DOI

Hongshan Guo, Dorit Aviv

thermal comfortcontrol

Energy and AI · 2026

Toward Smarter HVAC Control: Machine Learning Reveals Hidden Drivers in Thermal Comfort Databases.

Shows how missing-data policy reshapes feature sensitivity rankings and supports MRT-aware, occupant-centric HVAC control. The paper makes clear that preprocessing decisions materially affect what machine-learning models appear to learn from thermal comfort databases. DOI

Hongshan Guo, Ilaria Pigliautile, Yu Chang, Qingyao Qiao, Yichun Li

thermal comfortcontrolenergy
Browse all publications