EUVIMEDEUVIMED entdecken
Crossref

Hybrid physics-machine learning digital twin for high-resolution traction substation power estimation

Khoi Tran Van, Anh An Thi Hoai Thu

Transport and Communications Science Journal · 15.03.2026 · 77 (7) · 983-1000

DOI: 10.47869/tcsj.77.7.8 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Transport and Communications Science Journal, online erschienen am 15.03.2026. Verfasst von Khoi Tran Van und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Real-time estimation of traction power demand plays an important role in the operation and energy management of urban railway systems. This paper proposes a hybrid digital twin framework that integrates physics-based modeling with machine learning techniques to reconstruct real-time traction power profiles at traction substations using SCADA measurements of power and energy recorded at intervals of 3–5 minutes. The proposed digital twin framework consists of three main components. First, a physics-based load-flow model of the DC traction network is employed to ensure physical consistency in representing the electrical energy system of the railway line. Second, a residual learning module to identify the SCADA measurement instants within the physical model cycle, thereby enabling the prediction of substation power evolution in real time until the next measurement interval. Third, a post-prediction power correction module is implemented using an adaptive correction coefficient in order to minimize discrepancies between the model output and the actual system behavior. The proposed approach is validated using real operational data collected over several representative days from the Cat Linh – Ha Dong urban railway line. The verification results show that the proposed digital twin framework is capable of reconstructing high-fidelity real-time substation power profiles that closely follow the dynamic patterns observed in SCADA measurements. The framework successfully enforces strict energy conservation, driving the cumulative daily energy deviation to near zero, while maintaining robust generalization capability in tracking highly non-linear peak power demands throughout the entire operational duration across individual substations.

Zitieren

APA

Van, K. T., & Thu, A. A. T. H. (2026). Hybrid physics-machine learning digital twin for high-resolution traction substation power estimation. Transport and Communications Science Journal, 77(7), 983-1000. https://doi.org/10.47869/tcsj.77.7.8

Vancouver

Van KT, Thu AATH. Hybrid physics-machine learning digital twin for high-resolution traction substation power estimation. Transport and Communications Science Journal. 2026;77(7):983-1000. doi:10.47869/tcsj.77.7.8

BibTeX
@article{Van2026,
  author = {Van, Khoi Tran and Thu, Anh An Thi Hoai},
  title = {Hybrid physics-machine learning digital twin for high-resolution traction substation power estimation},
  journal = {Transport and Communications Science Journal},
  year = {2026},
  volume = {77},
  number = {7},
  pages = {983--1000},
  doi = {10.47869/tcsj.77.7.8}
}
Kostenlos starten →