BuildSTG: A Multi-building Energy Load Forecasting Method using Spatio-Temporal Graph Neural Network

2025-07-28

AI Summary

Existing data-driven approaches struggle to capture spatial dependencies among multiple buildings and shared energy consumption patterns. To address this, we propose a spatiotemporal graph neural network (ST-GNN) framework that constructs an adaptive graph structure by jointly encoding building-specific attributes and dynamic environmental factors. The framework integrates multi-level graph convolutions, spatiotemporal attention mechanisms, and an interpretable graph decomposition module to enable joint modeling and forecasting of building-level energy consumption. This design significantly enhances the model’s capacity to represent complex spatial correlations and heterogeneous energy usage behaviors. Extensive experiments on the Building Data Genome Project 2 dataset demonstrate that our method outperforms baseline models—including XGBoost, SVR, fully connected neural networks (FCNN), and GRU—in prediction accuracy, cross-building generalizability, and interpretability of the learned graph topology.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments

Abstract

Due to the extensive availability of operation data, data-driven methods show strong capabilities in predicting building energy loads. Buildings with similar features often share energy patterns, reflected by spatial dependencies in their operational data, which conventional prediction methods struggle to capture. To overcome this, we propose a multi-building prediction approach using spatio-temporal graph neural networks, comprising graph representation, graph learning, and interpretation. First, a graph is built based on building characteristics and environmental factors. Next, a multi-level graph convolutional architecture with attention is developed for energy prediction. Lastly, a method interpreting the optimized graph structure is introduced. Experiments on the Building Data Genome Project 2 dataset confirm superior performance over baselines such as XGBoost, SVR, FCNN, GRU, and Naive, highlighting the method's robustness, generalization, and interpretability in capturing meaningful building similarities and spatial relationships.

Problem

• Predicts multi-building energy loads using spatio-temporal data
• Captures spatial dependencies missed by traditional methods
• Improves accuracy and interpretability in energy forecasting

Innovation

• Spatio-temporal graph neural networks for forecasting
• Multi-level graph convolutional architecture with attention
• Interpretable graph structure for building similarities