An Artificial Intelligence Framework for Vegetation Identification in Overhead Distribution Networks using the IEC 61850 Standard

The identification of vegetation contact with the power grid is a critical issue for utilities, posing significant risks to the population. Undetected events can lead to system failures, such as short circuits, outages, cable degradation, and environmental hazards, including large-scale wildfires. Vegetation pruning near distribution networks is common but costly due to the difficulty of covering extensive areas. The low current magnitudes from these events, known as vegetation high-impedance faults, often require high-frequency components for detection. However, high-frequency signal acquisition technologies are expensive and impractical for widespread deployment, prompting alternative methods for identification and vegetation management. This study proposes a practical and reliable framework for detecting vegetation contact with the distribution grid. It employs IEC 61850 standard and extended time-window analysis, capturing up to one minute of neutral current signals. These signals are processed by an AI algorithm that classifies events and delivers results. The proposed framework was tested in one feeder on realtime simulation but it is scalable to all substation feeders using technology of centralized protection. The approach achieved 95% accuracy and a false positive rate below 1%.