CCAIS researchers present at BuildSys 2026

Three CCAIS researchers - Jan Buermann, Luke Nicholas and Connor Watson — travelled to Banff, Canada for ACM BuildSys 2026, the 13th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation (22–25 June 2026). Between them, they presented four pieces of work spanning home battery systems, the role of occupants in building energy use, and indoor temperature modelling — all aimed at making buildings more energy-efficient while considering the needs of people who use them.

Jan presented a full paper at the conference and Luke had a poster, which was also a finalist for the conference’s Best Poster Award! Luke and Connor both presented at the co-located OccuSys workshop on Occupant-Centric Energy Systems, which brings together researchers working on how to model and respond to building occupant behaviour.

The CCAIS papers and posters:

• Jan Buermann — Designing Personalised Residential Hybrid Battery Systems Through a Two-Tier Mixed-Integer and Genetic Algorithm Framework (BuildSys paper). A two-tier optimiser — a genetic algorithm paired with a mixed-integer scheduler — that designs personalised home battery systems combining technologies such as lithium-ion and lead-acid to make the most of rooftop solar.
• Luke Nicholas — Generative Occupants: LLM Agents in a Physics-Based Building Energy Simulation (poster; Best Poster Award finalist). Introduces “generative occupants” — large language model agents with memory and daily plans, coupled to a physics-based building simulation to capture how people’s behaviour shapes energy use.
• Luke Nicholas — Towards Generative Occupants: LLM Agents in Physics-Based Building Simulation (OccuSys workshop paper). The full workshop paper behind the poster, going into further details.
• Connor Watson — Simulation of Temperature in a Naturally Ventilated Office Space Using Machine Learning and Comparison with Traditional Whole-Building Temperature Modelling (OccuSys workshop paper). Compares a physics-informed machine-learning model (built with NVIDIA PhysicsNeMo) against a traditional whole-building energy model for predicting indoor temperature in a naturally ventilated office, benchmarked against real sensor data.