Data-Mining Techniques Based Relaying Support for Symmetric-Monopolar-Multi-Terminal VSC-HVDC System

Considering the advantage of ability data-mining techniques (DMTs) to detect and classify patterns, this paper explores their applicability for protection voltage source converter-based high direct current (VSC-HVDC) transmission systems. In spite location fault occurring points such as external/internal, rectifier-substation/inverter-substation, positive/negative pole DC line, stated approach is capable accurate detection, classification, location. Initially, local measurements at one end HVDC system are used in work extract feature vector. Once vector retrieved, DMTs trained tested identify types (internal faults, external AC faults) particular feeder. framework, several state-of-the-art machine learning (ML) models along with advanced deep (DL) model training testing. The proposed VSC-HVDC relaying comprehensively on a symmetric-monopolar-multi-terminal presents heartening results diverse operating conditions. show that studied belief network (DBN) based DL performs better compared other ML both classification accuracy DBN found be 98.9% noiseless condition 91.8% 20 dB noisy condition. Similarly, DBN-based DMT effective locations smaller percentage errors MSE: 2.116, RMSE: 1.4531, MAPE: 2.7047. This can an low-cost support tool system, it does not necessitate communication channel.