This article presents an accurate DC fault location method that applies parameter fitting. This technique first discusses the traveling wave (TW) propagation process in the decoupled line-mode network. We obtain the exact fault distance equation based on the analytical expressions for the wavefront of backward line-mode voltage TW. The adaptive multi-step Levenberg- Marquardt's (AMLM) algorithm is used for parameter fitting due to its fast processing speed and accuracy. The exact fault location can then be estimated using the parameter fitting results. The proposed fault location method is validated using a three-terminal HVDC system modeled on a real-time digital simulator (RTDS) platform. Based on the experimental results, the proposed method accurately detects the fault location, with all estimated errors smaller than 1%, and can withstand 40 dB noise interference. Moreover, the proposed method does not need a high sampling frequency and communication device. Its accuracy is independent of fault resistance and type compared to existing methods.@en

The existing protection techniques for high-voltage direct-current (HVDC) grids suffer from several shortcomings such as high sampling frequency, poor robustness, and reliance on simulation for threshold setting. To solve these problems, this paper proposes a non-unit protection method for modular multilevel converter (MMC)-based HVDC grids using the curvatures of backward traveling waves. To this end, the propagation characteristics of traveling waves and the boundary characteristics of DC lines are first studied, then the analytical expressions of backward traveling waves are derived. Moreover, the curvatures of backward traveling waves are analyzed. On this basis, a non-unit protection method is proposed, including zone selection, disturbance identification, and pole selection. At last, with a protection platform and a real-time digital simulator (RTDS) platform of the MMC-HVDC grid, the accuracy and the robustness of the proposed protection method are verified. The results show that the protection method can correctly identify faults with different distances and resistance in 1 ms and has strong robustness against transition resistance, sampling frequency, boundary value, noise, system topology, and line parameters.@en