Enhancing fuzzy cognitive map convergence through supervised and unsupervised learning algorithms: A case study of operational risk assessment in power distribution networks

Fuzzy Cognitive Maps (FCMs) are commonly used for modeling complex systems. However, their convergence challenges significantly limit their accuracy and reliability, especially in operational risk assessment. To address this issue, the current study proposes a novel approach that integrates advanced...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Engineering applications of artificial intelligence Jg. 161; S. 112104
Hauptverfasser: Baghemoortini, Elham Fallah, Shishebori, Davood, Jahangoshai Rezaee, Mustafa, Jabbarzadeh, Armin
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 09.12.2025
Schlagworte:
Convergence optimization
Fuzzy cognitive maps
Learning algorithms
Operational risk assessment
Power distribution network
Learning algorithms
Power distribution network
Operational risk assessment
Fuzzy cognitive maps
Convergence optimization
ISSN:0952-1976
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Fuzzy Cognitive Maps (FCMs) are commonly used for modeling complex systems. However, their convergence challenges significantly limit their accuracy and reliability, especially in operational risk assessment. To address this issue, the current study proposes a novel approach that integrates advanced supervised and unsupervised learning algorithms: specifically, the Mesh Adaptive Direct Search (MADS) and Genetic Algorithm (GA). To meet the critical need for accurate risk modeling in power distribution networks, the proposed methodology utilizes ten years of time-series data from the Yazd Power Distribution Network as a real-life case study. This optimizes risk relationships and reduces convergence errors. The main contributions of this research are: (1) the development of an integrated FCM-based framework that improves convergence stability and accuracy through advanced learning algorithms, (2) the demonstration of the effectiveness of MADS in achieving faster convergence and lower error rates compared to GA, and (3) the provision of a data-driven, scalable solution for risk prioritization and decision-making in complex and dynamic systems. The results show a significant decrease in convergence error from 0.126 to 0.005, allowing for more precise risk analysis and mitigation strategies. •A novel hybrid learning model for FCMs using MADS and GA is developed.•The model improves FCM convergence and stability in complex systems.•Real-world time-series data from power networks are used for validation.•The proposed MADS outperforms GA in optimizing high-dimensional, non-linear systems.•The approach enables dynamic and scalable risk assessment across domains.
ISSN:0952-1976
DOI:10.1016/j.engappai.2025.112104