Effect Of Conducting Particle On Spacers In SF6-N2 Gas Mixture Subjected To HVAC Using Physics-Informed Neural Network (PINN) Model And Hybrid Optimization Algorithm

Abstract

Insulation degradation is a major problem in the HV equipment and often causes the failure of equipment and power outages. Sulphur hexafluoride has high dielectric strength but also a high global warming potential, which means we still need an environmentally friendly alternative. A good alternative appears to be a gaseous mixture composed of 10% SF₆ and 90% N₂ for insulation preservation and environmental safety. This study investigates the breakdown voltage characteristics of solid dielectrics, specifically Polymethyl Methacrylate (PMMA) and Nylon, under high-voltage alternating current (HVAC) stress in the SF₆–N₂ environment, considering the effect of conducting particles. To enhance prediction accuracy and system intelligence, three artificial intelligence (AI)-based methodologies are integrated. A Physics-Informed Neural Network (PINN) model is implemented to embed Maxwell’s equations directly into the neural network, enabling physically consistent predictions. Additionally, a hybrid optimization algorithm combining the Improved Salp Swarm Algorithm (ISSA) and Adaptive Differential Evolution (ADE) is employed to optimize spacer geometry for reduced electric field enhancement. A Graph Neural Network (GNN) is used to forecast electric field intensities and locate potential failure zones. This integrated approach significantly improves the reliability of HV insulation systems.