An online learning method for assessing smart grid stability under dynamic perturbations
Abstract The increasing complexity of smart grid (SG) systems necessitates advanced methodologies to ensure their stability and reliability. In this work, we propose a novel online learning framework that leverages the Bee Algorithm for Ensemble Learning (BAEL) with dynamic perturbations to enhance...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-03-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-94718-3 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract The increasing complexity of smart grid (SG) systems necessitates advanced methodologies to ensure their stability and reliability. In this work, we propose a novel online learning framework that leverages the Bee Algorithm for Ensemble Learning (BAEL) with dynamic perturbations to enhance the adaptability and performance of ML models in SG stability prediction. The key contributions of our approach are twofold. First, we introduce a dynamic perturbations mechanism that systematically adjusts variations within the Bee Algorithm, effectively balancing global exploration speed and local convergence accuracy throughout the learning process. Second, we integrate the BAEL strategy, where model selection and evolution are guided by performance-driven ensemble learning, allowing continuous adaptation to evolving data patterns. Through iterative learning cycles augmented by incremental perturbation adjustments, our method significantly improves predictive accuracy. To evaluate the effectiveness of our approach, we conduct extensive experimental assessments, demonstrating that our online learning process achieves an F1-score close to 100 percent. Additionally, we perform a comparative analysis between the Bee Algorithm and a benchmark fusion model incorporating Random Forest (RF), Gradient Boosting (GB), and eXtreme Gradient Boosting (XGB) classifiers, under identical conditions, including the presence of dynamic perturbations. The results confirm that our BAEL-based approach consistently outperforms both the fusion of these classifiers and each of them operating independently across all evaluation metrics, highlighting its robustness in predicting SG stability under dynamic perturbations. |
|---|---|
| ISSN: | 2045-2322 |