Density-Based Adaptive Protection for HVDC Transmission Lines Using Statistical Learning and Proactive Control

To address the challenges in the reliability of high-voltage direct current (HVDC) transmission line protection under high-impedance fault conditions, this paper proposes a density-based adaptive protection scheme that integrates statistical learning and proactive control. First, principal component analysis (PCA) is employed to project fault features into a reduced-dimensional space. Then, kernel density estimation (KDE) is applied to determine adaptive protection boundaries. The corresponding protection strategies (e.g., rapid isolation, collaborative discrimination, or blocking) are activated based on the relative position between the real-time fault features and the density-based boundaries. The protection performance is evaluated by calculating the deviation between the projected features of actual faults and the KDE-derived boundaries. An adaptive mechanism continuously optimizes both the boundaries and strategies, ensuring reliable performance under various operating conditions. Validation via electromagnetic transient simulations and field data confirms the scheme's accurate fault identification and operational reliability under different fault types, transition resistances, and noise conditions.