Building simulation based on physical modeling is commonly adopted for performance prediction, but the high time costs hinder its application in the early design stage of buildings. Data-driven surrogate models have been proposed as a means to replicate computationally expensive simulation models. However, existing surrogate models for sustainable residential block design are limited in scope, focusing either on individual buildings or on specific cases within multi-block projects. This study leverages graph neural networks to develop optimal surrogate models incorporating inter-building effects to predict multiple indicators of sustainable performance for residential blocks at a region level. A graph schema is prop osed to represent the general geometric features and relations among buildings in residential bloc k design. A regional dataset is generated for model training and testing, using real residential zones in Hong Kong.
| Authors | Wu, Zhaoji and Li, Mingkai and Liu, Wenli and Cheng, Jack C. P. and Wang, Zhe and Kwok, Helen H. L. and Huang, Cong and Hou, Fangli |
|---|---|
| Year | 2025 |
| Venue | Building Simulation: An International Journal |
| DOI | 10.1007/s12273-025-1237-7 |
| Source Database | EBSCO (Springer Nature Journals) |
| Bridge-to-GNN | Category A |
| GNN Architecture | GAT (heuristic — verify) |
| Graph Encoding | Not yet extracted from PDF |
| AEC Task | Not yet extracted from PDF |
| Cohort | Mature Applications (2024–2026) |
| Dataset Size | 962 graphs; 26 buildings (heuristic — verify) |
|---|---|
| Implementation Framework | PyTorch Geometric (PyG), PyTorch (heuristic — verify) |
| Key Hyperparameters | 2000 epochs (heuristic — verify) |
| Primary Metric | Accuracy (heuristic — verify) |
| Primary Value | 11.79% (heuristic — verify) |
| Key Finding | graph schema describing general residential block design, which supports various input structures of the GNN-based surrogate models and represents inter-building effects (heuristic — verify) |
| Quality Assessment | 1/6 · rigor: Low · Baseline: ✓ (heuristic — verify) |
| GNN architectures detected | GAT (Graph Attention) · T-GCN / Temporal GCN · Knowledge Graph Embedding (TransE/RotatE/DistMult) · GCN |
|---|---|
| Frameworks / libraries | PyTorch Geometric (PyG) · PyTorch · EnergyPlus |
| Hyperparameters (regex-detected) | epochs=2000 |
| Dataset stats found | 26 buildings · 962 graphs |
| Reported metrics + values | Accuracy: 11.79% · Accuracy: 12.9% · Accuracy: 10.2% · RMSE: 11.79% · RMSE: 12.05% · RMSE: 0.5% · RMSE: 3.2% · RMSE: 21.9% |
| Quality (heuristic, 6-flag) | 1/6 · rigor: Low Code/Data: —Benchmark: —Baseline: ✓CV/Split: —Ablation: —Reproducible: — |
Part of the GML/GNN in AEC Systematic Review (PRISMA 2020) — 112 papers, 2020–2026