Fault Detection, Faulted-Line Identification and Fault Location in a Three-Terminal VSC-HVDC System with Hybrid Cable-Overhead Lines

Protection schemes for DC faults in multi-terminal VSC-based HVDC systems (MTDC) present a challenge as system expansion allows a single terminal to connect multiple DC transmission lines. A DC fault in an HVDC system is characterized by a fast rise of current, with the fault propagates to all interconnected converter stations. This necessitates protection schemes capable of detecting a DC fault within a few milliseconds to prevent damage, given the VSC's vulnerability to high currents. Furthermore, the scheme must identify which DC line is faulted and selectively isolate only the faulted line. Once the faulted line is isolated, an important subsequent task is to estimate the fault location along the line, which provides preliminary information on the distance to the corresponding terminal. Using the recorded waveforms of both line-end measurements, offline fault location is the next stage that complements the protection schemes by extracting waveform features to determine the fault's position. This thesis combined the protection scheme and post-fault location analysis for a three-terminal VSC-based HVDC system, studied in PSCAD simulation. The system is configured as a radial network with one uniform line (overhead line) and one hybrid line (a mix of cable-overhead line). A fault detection method based on single-ended measurements is used for the protection schemes in the system model. The rate-of-change-of-current (ROCOC) is utilized as the algorithm can detect and achieve the selectivity by a predefined threshold. The fault location estimation then proceeds using the Maximum Overlap Discrete Wavelet Transform (MODWT) and a supervised learning technique of Artificial Neural Networks (ANN). These two methods present a learned estimator based on feature-to-distance mapping. Fault cases are generated by varying fault resistance, fault type (pole-to-pole and pole-to-ground), and fault location within the DC transmission line. The result demonstrates that the proposed protection scheme can selectively isolate the majority of fault cases. The offline fault-location stage estimates the fault position with a mean absolute error below 0.4 km, corresponding to a mean relative error of less than 0.92\% of the fault distance.