Free‐Air Electric Arc Phenomena in High‐Voltage Energy Transmission Systems: A Comprehensive Review Integrating Physical Modeling and Intelligent Monitoring

ABSTRACT Reliable detection of free‐air electric arc faults in high‐voltage overhead transmission lines, together with accurate discrimination between transient and permanent faults, is essential for secure single‐phase auto‐reclosing and for maintaining system stability. However, the nonlinear, time‐varying, and environmentally dependent nature of secondary arcs, as well as the influence of shunt reactors, line configuration, and measurement constraints, has led to a fragmented body of literature with no unified analytical map of available detection strategies. This paper presents a systematic and analytical review of the reported methods for free‐air arc‐fault detection in overhead transmission lines. The reviewed studies are classified into three main categories: circuit‐equation‐based methods, signal‐processing‐based methods, and pattern‐recognition/intelligent‐learning‐based methods. For each category, the conceptual basis, required measurements, processing sequence, decision indices, implementation steps, and numerical or experimental limitations are extracted and synthesized. In addition, representative methods are comparatively evaluated using an integrated set of technical and operational criteria, including detection accuracy, response speed, robustness to noise and operating‐condition variation, information requirements, computational burden, and real‐time implementation capability. The results show a clear methodological evolution from physics‐based analytical formulations toward data‐driven and adaptive detection frameworks. Circuit‐equation‐based approaches provide high physical interpretability and direct estimation of arc‐related parameters, but their performance is often constrained by model assumptions, threshold dependence, and multi‐terminal measurement requirements. Signal‐processing‐based methods offer faster and more localized detection by exploiting transient, harmonic, and time–frequency signatures of the arc, although their stability and generalizability may be affected by noise level, parameter tuning, and signal decomposition settings. Intelligent‐learning‐based methods achieve the highest adaptability and classification capability in complex conditions, but they remain strongly dependent on the quality and diversity of training datasets and impose greater computational and implementation burdens. Overall, the review indicates that no single methodological family simultaneously satisfies the full set of protection requirements for modern transmission systems. The main conclusion is that the most promising direction for next‐generation arc‐fault detection lies in hybrid and multi‐layer architectures that combine physical modeling, transient signal analysis, and adaptive intelligent decision‐making within relay‐deployable protection frameworks.