ArchEGraph Dataset Links Building Geometry to Energy Performance

Yihui Li, Yihui Chen, Kaidi Zha, Xiaoyue Yan, Zhexuan Yu, Shiqi Dai, Jun Xiao, Jun Yin, Ramon Elias Weber, Borong Lin· August 10, 2026 View original

Key takeaways

  • ArchEGraph is a large-scale dataset for building energy modeling using heterogeneous graphs.
  • It aligns building geometry, topology, weather, and thermal loads.
  • The dataset supports tasks like graph reconstruction and topology-informed load prediction.
  • It enables the development of scalable and generalizable surrogate models for sustainable design.

Who benefits

ArchitectureConstructionUrban PlanningEnergy ManagementReal Estate

Summary

ArchEGraph is a new large-scale dataset representing buildings as heterogeneous graphs, aligning geometry, topology, weather, and thermal loads. It aims to improve building energy modeling and accelerate the development of machine learning models for sustainable design.

This paper introduces ArchEGraph, a significant new benchmark dataset designed to advance building energy modeling and support the development of machine learning models for sustainable architecture. The dataset uniquely represents buildings as heterogeneous graphs, meticulously aligning their geometric and topological structures with weather conditions and zone-level thermal loads. Comprising 5,481 buildings and 49,326 validated building-weather simulation cases, ArchEGraph offers substantial complexity with over 133,000 space nodes and 1.44 million face nodes. The researchers define two key benchmark tasks: graph reconstruction from polygonal meshes to recover topological structure, and topology-informed load prediction, which forecasts zone-level thermal response using graph structure and temporal weather data. Standardized evaluation protocols are provided, along with cross-building and cross-climate generalization experiments to assess model robustness.

Why it matters

Professionals in architecture, engineering, and urban planning can use this dataset to develop more accurate and rapid AI-driven tools for designing energy-efficient buildings, contributing significantly to carbon neutrality goals.

How to implement this in your domain

  1. 1Access and utilize the ArchEGraph dataset to train and validate machine learning models for building energy prediction.
  2. 2Develop novel graph neural network (GNN) architectures specifically tailored for the geometry-topology-physics coupling in buildings.
  3. 3Integrate insights from topology-informed load prediction into early-stage building design workflows.
  4. 4Collaborate with researchers to extend the dataset or develop new benchmark tasks for building performance optimization.

Original post by Yihui Li, Yihui Chen, Kaidi Zha, Xiaoyue Yan, Zhexuan Yu, Shiqi Dai, Jun Xiao, Jun Yin, Ramon Elias Weber, Borong Lin

"arXiv:2608.06772v1 Announce Type: new Abstract: Accurate estimation of building energy use is essential for achieving carbon neutral and sustainable buildings. To better understand the influence of design decisions on building energy use and calibrate machine learning models that…"

View on X

Originally posted by Yihui Li, Yihui Chen, Kaidi Zha, Xiaoyue Yan, Zhexuan Yu, Shiqi Dai, Jun Xiao, Jun Yin, Ramon Elias Weber, Borong Lin on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses