Data-Driven Approaches to Reliability Modeling in Critical Energy Infrastructure: Analytical Perspectives on Early Failure Detection

Authors: Derrick Ohene Adusei

DOI: https://doi.org/10.37082/IJIRMPS.v14.i2.233031

Short DOI: https://doi.org/hbvqn8

Country: United States

Full-text Research PDF File:   View   |   Download

Abstract: The growing complexity and interconnection of critical energy infrastructure, which stems out of integration of renewables, decentralization of generation, and cyber-physical integration, has increased the pressure on the sophisticated reliability modeling and early warning of failure systems. Although traditional probabilistic methods are grounded, they have disadvantages in terms of being state-based and less adaptive to high-dimensional operation data. This paper is a PRISMA systematic literature review of data-driven reliability models in critical energy infrastructure based on publications in ScienceDirect, having selected on the use of structured Boolean search operators. Eighty-eight studies were retained out of an original 296 records after a process of removal of duplicates and systematic screening. The review constructs a taxonomy of statistical, machine learning and deep learning, and hybrid physics-informed models, and assesses their applicability in early failures estimation of transformers, wind turbines, photovoltaic systems, nuclear systems as well as smart grids. A comparative analysis shows increasing popularity of ensemble and deep learning architectures, specifically, to anomaly detection and remaining useful life (RUL) prediction. Nevertheless, there are chronic problems such as rare-event imbalance, benchmarking fragmentation, limited cross-domain transferability, interpretability limitations as well as a lack of validation on real-world operational data that limit cumulative improvements. The paper lists such strategic research priorities as the integration of physics-informed AI, federated and edge learning models, frameworks of transfer learning, unlabelled data self-supervised learning, and standardized benchmarking data. The results give an all-purpose analytical starting point in the Standard of predictive, scalable, and robust early failure detection systems in the future energy infrastructure.

Keywords: Data-driven reliability modelling, Early failure detection, Predictive maintenance, Critical energy infrastructure, Physics-informed artificial intelligence

Paper Id: 233031

Published On: 2026-04-02

Published In: Volume 14, Issue 2, March-April 2026

Share this