Research output
Selected work since joining HKU in September 2023: journal articles and refereed conference and workshop papers, including accepted manuscripts.
Select multiple areas to find work at their intersection.
Showing all 21 papers
No publications match this combination. Try removing an area or topic, or .
3 papers
Energy and Buildings
Accepted 27 September 2026.
Uses predicted probabilities of thermal sensation to inform HVAC supervision, examining delivered-energy and operative-temperature trade-offs across a weather stress grid.
Engineering Applications of Artificial Intelligence
Accepted 15 September 2026.
Energy [DOI ↗]
In press
Article 142403; DOI registered, volume not yet assigned.
Examines how future weather and population ageing affect residential cooling service and thermal outcomes under equal cooling setpoints.
12 papers
Building and Environment [DOI ↗]
Published online
Combines dense sensor measurements with 13 months of coffee-shop and restaurant monitoring to identify CO₂ build-up and recovery patterns. Recent occupancy history improves prediction of concentration changes, while the diagnostic framework distinguishes temporal response from instantaneous headcount or ventilation estimates.
Sustainable Cities and Society [DOI ↗]
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.
Energy and Buildings [DOI ↗]
Develops a stochastic weather-generation workflow using multi-decadal observations to support scenario-conditioned building performance analysis under climate uncertainty.
Energy and Buildings [DOI ↗]
Finds that gender differences become most pronounced under high-velocity cooling at lower temperatures, with implications for equitable fan-cooling design. By connecting subjective votes with skin-temperature responses, the study shows why mixed-mode cooling strategies should not assume uniform comfort response.
Energy [DOI ↗]
Shows that weather-file baseline quality dominates future demand bias, outperforming more complex workflows for climate-adapted energy projections. Rather than rewarding statistical complexity for its own sake, the paper identifies input fidelity as the main determinant of robust downstream building simulation.
Building and Environment [DOI ↗]
Predicts the probability of each response on the seven-point cold-to-hot thermal sensation scale, so design and control decisions can account for uncertainty in how occupants feel.
Energy and AI [DOI ↗]
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.
Energy and Buildings [DOI ↗]
Combines field diagnostics and cross-climate simulation to quantify ventilation, comfort, and energy trade-offs in retrofitted Hong Kong wet markets. It frames wet-market modernization as both a building-performance problem and a public-institution design question.
Energy and Buildings [DOI ↗]
Integrates data-driven thermal sensation models with building control workflows to benchmark occupant-aware control across multiple climates. It shows how co-simulation can move comfort models from offline prediction into actionable control testing.
Energy and Buildings [DOI ↗]
Demonstrates that demographic-aware metabolic loads can cut HVAC energy use and reduce gender-based comfort bias relative to the legacy 120 W/person assumption. The work directly challenged a century-old default embedded in standards and helped motivate ASHRAE 1959-TRP.
Renewable and Sustainable Energy Reviews [DOI ↗]
Reviews how comfort studies use personal, contextual, and PMV-related variables, highlighting gaps that limit cross-cultural comparison and model transfer. It also establishes a benchmark-oriented framing for comparing Eastern and Western comfort evidence more systematically.
Building and Environment [DOI ↗]
Uses physiological constraints inside a neural model to improve robustness and interpretability in large-scale thermal comfort prediction. The approach treats building-comfort ML as a physically informed modeling problem rather than a purely statistical fitting exercise.
6 papers
E3S Web of Conferences 716 [DOI ↗]
Best Paper Award
IAQVEC 2026, Los Angeles, May 2026. Future-compatible climate zoning for building codes.
ICML 2025 CO-BUILD Workshop
Oral, first place
Smart Building Competition winner.
ICML 2025 CO-BUILD Workshop
Poster, third place
Recognized at the NeurIPS 2025 Urban AI Workshop Buildings Challenge.
ICML 2025 CO-BUILD Workshop
Poster
Multiphysics and Multiscale Building Physics (IBPC 2024) [DOI ↗]
Proceedings chapter, pp. 55–60; first published online in December 2024.
Multiphysics and Multiscale Building Physics (IBPC 2024) [DOI ↗]
Proceedings chapter, pp. 585–591; first published online in December 2024.