University of Illinois Urbana-Champaign

Grey-Box Digital Twins of Nuclear Power Plants

Miqueles, Leonardo; Ahmed, Ibrahim; Di Maio, Francesco; Zio, Enrico

Content Files
A70.pdf

Permalink

https://hdl.handle.net/2142/121859

Description

Title
Grey-Box Digital Twins of Nuclear Power Plants
Author(s)
  • Miqueles, Leonardo
  • Ahmed, Ibrahim
  • Di Maio, Francesco
  • Zio, Enrico
Issue Date
2023
Keyword(s)
  • Nuclear power plant
  • Digital twin
  • Grey-box modeling
  • Dual fluid reactor
Abstract
Digital Twins (DTs) can enable risk monitoring and control of Nuclear Power Plants (NPPs). The underlying concept is to integrate the informative data measured on the Physical Object (PO) with that generated by its digital counterpart, known as the Digital Object (DO), which mimics its behaviour. For computational feasibility, DTs usually rely on data-driven Black-Box (BB) models that are trained on historical data collected during past PO operations. The lack of interpretability of the input-output relationships embedded in the BB models make it difficult to use DTs in a safety- critical industrial sector like the nuclear one. To address this issue, we propose to use a Grey-Box (GB) modelling approach for DT development, trading-off accuracy, computational burden and interpretability. The proposed GB approach is exemplified on a case study concerning a small modular dual fluid reactor (SMDFR). A dynamic model of the SMDFR is taken as PO and is used to simulate accidental scenarios by a MC-driven fault injection method. The GB DO is made of physics-based White-Box (WB) and data-driven BB models for virtual sensing non-measurable key process variables. A real-time risk monitoring module is, then, built to provide probabilistic safety margins (PSMs) of relevant safety variables. A comparative analysis of GB configurations is presented and methodological challenges are discussed. The results of the case study show the advantages of the proposed GB DT-based method over purely WB and BB modelling approaches.
Type of Resource
text
Genre of Resource
Conference Paper/ Presentation
Language
eng
Handle URL
https://hdl.handle.net/2142/121859

Owning Collections

PSAM 2023 Conference Proceedings PRIMARY