Article: Research on a collaborative calculation framework for cross-regional power grid carbon emissions based on federated learning and adaptive graph convolution Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.50 pp.70 - 94 Abstract: This paper proposes a federated learning framework integrated with adaptive graph convolution for accurate and privacy-preserving carbon emission calculation in cross-regional power grids. It addresses data silos and privacy concerns by training models locally, avoiding raw data transfer. The adaptive graph convolution component automatically captures the dynamic spatial dependencies and carbon flow effects between grid regions. Validated on a Chinese grid dataset, the method reduces calculation errors by 22.3% and 14.7% compared to centralised and traditional distributed approaches, respectively, while demonstrating strong robustness against grid topology and operational fluctuations. Inderscience Publishers - linking academia, business and industry through research

Title: Research on a collaborative calculation framework for cross-regional power grid carbon emissions based on federated learning and adaptive graph convolution

Authors: Zhiqi Chen; Sheng Chen; Gang Yao; Kang Li; Honglue Zhang; Yang Wu; Mingyang Wang

Addresses: Power Dispatch and Control Center, Guizhou Power Grid Co., Ltd., Guiyang, 550002, China ' Power Dispatch and Control Center, Guizhou Power Grid Co., Ltd., Guiyang, 550002, China ' Power Dispatch and Control Center, Guizhou Power Grid Co., Ltd., Guiyang, 550002, China ' Power Dispatch and Control Center, Anshun Power Supply Bureau, Guizhou Power Grid Co., Ltd., Anshun, 561000, China ' Power Dispatch and Control Center, Guizhou Power Grid Co., Ltd., Guiyang, 550002, China ' Power Dispatch and Control Center, Guizhou Power Grid Co., Ltd., Guiyang, 550002, China ' Power Dispatch and Control Center, Guizhou Power Grid Co., Ltd., Guiyang, 550002, China

Abstract: This paper proposes a federated learning framework integrated with adaptive graph convolution for accurate and privacy-preserving carbon emission calculation in cross-regional power grids. It addresses data silos and privacy concerns by training models locally, avoiding raw data transfer. The adaptive graph convolution component automatically captures the dynamic spatial dependencies and carbon flow effects between grid regions. Validated on a Chinese grid dataset, the method reduces calculation errors by 22.3% and 14.7% compared to centralised and traditional distributed approaches, respectively, while demonstrating strong robustness against grid topology and operational fluctuations.

Keywords: federated learning; adaptive graph convolutional networks; grid carbon emissions; collaborative computing; privacy protection.

DOI: 10.1504/IJICT.2026.153627

International Journal of Information and Communication Technology, 2026 Vol.27 No.50, pp.70 - 94

Received: 17 Nov 2025
Accepted: 01 Jan 2026
Published online: 18 May 2026 *

Title: Research on a collaborative calculation framework for cross-regional power grid carbon emissions based on federated learning and adaptive graph convolution

Keep up-to-date