<![CDATA[Purpose – This study aims to develop an uncertainty-aware decision-support framework for prioritizing anomaly events in smart grid systems, addressing the limitation of existing approaches that focus primarily on binary anomaly detection rather than operational risk-based prioritization. Design/methods/approach – A hybrid framework is proposed by integrating lightweight statistical anomaly screening with a Mamdani-type fuzzy inference system (FIS) that models three risk dimensions: Impact, Likelihood, and Criticality. To enhance ranking performance while maintaining interpretability, fuzzy membership parameters and rule weights are optimized using a genetic algorithm (GA). The model is evaluated using the Open Power System Data dataset with 5-fold cross-validation. Findings - Experimental results show that the proposed model outperforms baseline methods, achieving NDCG@10 of 0.84, Recall@5 of 0.81, and F1-score of 0.86. The framework also reduces false alarm rates and maintains end-to-end latency below 300 ms on ARM-class hardware, demonstrating suitability for real-time edge deployment. Research implications/limitations – The study is limited to a single regional dataset and relies on offline optimization of fuzzy parameters. Future research may explore multi-regional validation, online adaptive learning, and integration with cyber-physical anomaly streams. Originality/value – This study introduces a novel hybrid framework that combines statistical screening, interpretable fuzzy risk modeling, and evolutionary optimization for anomaly prioritization. Unlike conventional detection-focused approaches, it emphasizes ranking quality, uncertainty handling, and real-time embedded feasibility, offering a practical and explainable solution for smart grid operations.]]>
Copyrights © 2026
