Year: 2024
Type: article
Abstract: Machine Learning-based fault detection approaches in energy systems have gained prominence for their superior performance. These automated approaches can assist operators by highlighting anomalies and faults, providing a robust framework for improving Situation Awareness. However, existing approaches predominantly rely on monolithic models, which struggle with adapting to changing data, handling imbalanced datasets, and capturing patterns in noisy environments. To overcome these challenges, this study explores the potential of Multiple Classifier System (MCS) approaches. The results demonstrate that ensemble methods generally outperform single models, with dynamic approaches like META-DES showing remarkable resilience to noise. These findings highlight the importance of model diversity and ensemble strategies in improving fault classification accuracy under real-world, noisy conditions. This research emphasizes the potential of MCS techniques as a robust solution for enhancing the reliability of fault detection systems.
Source: Energies
DOI: 10.3390/en17225787
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,955
Award: G359972691
OpenAlex ID: W4404546580