Article: Topology-aware modelling of regional energy systems using hierarchical graph networks for carbon peaking analysis Journal: International Journal of Global Warming (IJGW) 2026 Vol.39 No.5 pp.1 - 10 Abstract: A physics-constrained hierarchical graph network is developed for carbon forecasting based on directed energy-flow graphs. The model combines topology-specific attention with temporal encoding, hierarchical pooling and an energy-balance penalty to capture interregional flow dependence and thermodynamic consistency. Using 30-province data from 2000-2022, it achieves a root mean square error of 14.32 million tonnes of CO<SUB align="right"><SMALL>2</SMALL></SUB>, 27.86% lower than the strongest baseline, a coefficient of determination of 0.98, and an energy imbalance rate of 0.42%. Attribution analysis assigns 42.00% of emission growth to coal-to-power flows. Under Shared Socioeconomic Pathway 2, Jiangsu emissions decline after 2020, whereas Inner Mongolia peaks in 2029. Inderscience Publishers - linking academia, business and industry through research

Title: Topology-aware modelling of regional energy systems using hierarchical graph networks for carbon peaking analysis

Authors: Yun Zhang

Addresses: School of Economics and Management, Zhengzhou Normal University, Zhengzhou, Henan, 450044, China

Abstract: A physics-constrained hierarchical graph network is developed for carbon forecasting based on directed energy-flow graphs. The model combines topology-specific attention with temporal encoding, hierarchical pooling and an energy-balance penalty to capture interregional flow dependence and thermodynamic consistency. Using 30-province data from 2000-2022, it achieves a root mean square error of 14.32 million tonnes of CO₂, 27.86% lower than the strongest baseline, a coefficient of determination of 0.98, and an energy imbalance rate of 0.42%. Attribution analysis assigns 42.00% of emission growth to coal-to-power flows. Under Shared Socioeconomic Pathway 2, Jiangsu emissions decline after 2020, whereas Inner Mongolia peaks in 2029.

Keywords: graph neural networks; regional energy systems; energy-flow graph; carbon peaking; physics-constrained learning; dynamic attribution.

DOI: 10.1504/IJGW.2026.155790

International Journal of Global Warming, 2026 Vol.39 No.5, pp.1 - 10

Received: 21 Jan 2026
Accepted: 30 Jun 2026
Published online: 13 Aug 2026 *

Title: Topology-aware modelling of regional energy systems using hierarchical graph networks for carbon peaking analysis

Keep up-to-date