Monitoring Leakage Current in Insulator Strings for Flashover Risk Prediction: A Systematic Literature Review

The accumulation of pollution on high-voltage insulators, combined with humidity, forms a conductive layer that drives leakage current (LC) activity, potentially leading to critical flashover events. This work presents a Systematic Literature Review (SLR) designed not only to map the current state of sensing technologies but to characterize the ongoing technological disruption in predictive maintenance. Analyzing 31 primary studies selected from 3939 records, this review identifies a paradigm shift from traditional statistical indicators to advanced Deep Learning (DL) architectures. Quantitative analysis reveals that modern pulse-based monitoring and DL approaches achieve prediction accuracies exceeding 98% in real-time scenarios, while hybrid multimodal frameworks (e.g., CNN-LSTM fused with climatic data) reach up to 99.2% in risk classification. Beyond synthesizing current benchmarks, this work contributes by translating research gaps into concrete technical directions for next-generation systems. These proposed directions include: (1) the application ofGenerative AI (GANs and Diffusion Models)to synthesize rare fault data and mitigate class imbalance; (2) the transition from standard RNNs toTransformers and xLSTMarchitectures for capturing long-term pollution trends; (3) the deployment ofTinyMLfor edge-intelligence and online learning; and (4) the use ofGraph Neural Networks (GNNs)for optimizing sensor placement in transmission topologies. This study serves as a technical roadmap for implementing resilient, AI-driven insulator monitoring systems.