Evolution of Voltage Source Converter‐Based High‐Voltage Direct Current Protection Technologies and Emerging Intelligence Trends Enabled by Large Language Models

Abstract

The accelerating transition toward carbon‐neutral energy systems drives the large‐scale integration of renewable generation, reshaping modern power grids. As a key enabler of long‐distance renewable energy transmission, Voltage Source Converter‐based High‐Voltage Direct Current systems have become essential infrastructure. However, line protection remains a critical bottleneck due to ultrafast fault transients, strong converter–line coupling, and sensitivity under weak‐grid conditions. This review identifies three intrinsic paradoxes constraining existing methods: the speed–reliability trade‐off, the laboratory–field generalization gap, and the intelligence–explainability tension. Conventional traveling‐wave and differential schemes struggle to maintain robustness in multi‐terminal and renewable‐dominated scenarios. Data‐driven approaches, including machine learning and deep learning, improve adaptability but rely on statistical correlations, leading to limited interpretability and reduced robustness under unseen conditions. To address these challenges, this review discusses an Large Language Model–Knowledge Graph‐based decision‐support paradigm that supports protection supervision, logic validation, adaptive setting review, and post‐fault diagnosis, while retaining primary trip initiation within deterministic protection. By combining mechanism‐oriented representation with semantic reasoning, this framework supports interpretable, consistent, and adaptive supervisory protection functions, paving the way for next‐generation DC systems in renewable‐dominated grids.