Enhanced Prediction of Insulator Flashover Voltage Using Artificial Neural Networks Optimized with Particle Swarm Optimization
Résumé
Outdoor insulators are critical components in power systems but are highly susceptible to environmental factors such as moisture, rain, and contaminants. These adverse conditions often lead to surface flashovers, causing insulation failures and compromising the reliability of power systems. To address this challenge, this study aimed to develop a robust predictive model for flashover voltage under polluted conditions, leveraging the capabilities of Artificial Neural Networks (ANN) optimized with Particle Swarm Optimization (PSO). The primary objective was to enhance prediction accuracy and overcome the limitations of traditional methods. The proposed ANN-PSO model was trained and validated using a comprehensive dataset comprising experimental and simulated data under various pollution conditions. Key input features included Equivalent Salt Deposit Density (ESDD), insulator height, leakage distance, diameter, form factor, and environmental conditions. PSO was employed to optimize the ANN parameters, minimizing error metrics, such as Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE), while maximizing the regression coefficient (R²). The results demonstrate that the proposed ANN-PSO model significantly outperforms traditional approaches, including standard ANN and hybrid models such as LS-SVM-PSO and LS-SVM-GWO, in terms of predictive performance. The model achieved exceptionally low values for RMSE (0.003511) and MAPE (0.842%) along with a very high R² value (0.997), confirming its precision, robustness, and superior capability to predict flashover voltage accurately. This study provides a practical and reliable tool for power utilities to monitor and mitigate the risk of insulator flashovers in polluted environments. Furthermore, it highlights the potential for integrating advanced hybrid AI models to address complex challenges in power system operations, paving the way for further innovation in predictive modeling for power system reliability.
Citer ce document
Accès au document
Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter
Voir l'article sur le site de la revueStatistiques
Consultations : 1
Téléchargements : 0