A New Accurate and Robust Method Based on Deep Neural Networks for Fault Location in HVDC Systems

This paper proposes a new method based on Deep Neural Networks and direct-current (DC) voltage signal for fault location in High Voltage Direct Current (HVDC) systems. The proposed solution uses only local measurements and a reasonable sampling frequency, which ensures its applicability. Performing an intelligent signal processing, the power system reliability and availability can be improved, without additional devices or communication links. The proposed method takes advantage of the high capability of neural networks for pattern recognition, associating the frequency spectrum of faulted signals to specific fault locations. The training process is detailed and the definition of the most suited neural network architecture is presented. The method is validated by using different HVDC systems, different fault types and fault resistances. The results reveal the good performance and robustness of the proposed method and its potential for application for real scenarios.