Intelligent fault diagnosis and protection strategy for flexible HVDC transmission systems

Abstract

Flexible high-voltage direct current (HVDC) transmission based on voltage-source converters (VSCs) and modular multilevel converters (MMCs) is becoming a critical interface for renewable-energy delivery, port and rail energy facilities, and high-power charging infrastructure. From an automation-control perspective, the protection layer must convert synchronized measurements into selective control commands within only a few milliseconds while avoiding misoperation under noise, high fault resistance, and communication latency. This paper proposes a dual-domain attention diagnosis and adaptive protection logic (DDA-APL) framework for flexible HVDC systems. The method fuses optical current/voltage measurements, traveling-wave features, pole voltage/current transients, and converter-state variables. A one-dimensional convolutional network extracts steep-front fault components, a gated temporal convolution layer preserves short-window dynamics, and a self-attention module assigns adaptive weights to informative channels. A multitask head outputs the fault category, line zone, distance estimate, and protection confidence. These outputs drive an action-scoring controller that coordinates selective direct-current (DC) circuit-breaker tripping, converter blocking, current-limiter insertion, and submodule bypass while retaining a deterministic backup path. A ±500kV, 1000 MW MMC-based HVDC benchmark was built in electromagnetic transient simulation, and 7200 cases were generated across fault types, locations, resistances, noise levels, and operation modes. The proposed method achieved 99.2% classification accuracy, a 0.72km mean location error, and a 1.38ms median protection decision time, outperforming the traveling-wave threshold, support vector machine (SVM), random forest, and convolutional neural network-long short-term memory (CNN-LSTM) baselines. The results demonstrate that the closed-loop integration of intelligent diagnosis and protection constraints improves sensitivity, security, and autonomous control performance in converter-station protection.