Mathematical modelling and data-driven protection algorithm for fault current calculation in bipolar medium voltage multi-terminal DC networks

초록

Faults in Medium Voltage Direct Current (MVDC) distribution networks present unique challenges, particularly in systems with bipolar configurations. The critical challenges include the high-speed fault current rise due to low network impedance, making fault interruption difficult, and the complex coordination of protection schemes in multi-terminal configurations. This paper focuses on the calculation of fault currents in a five-terminal bipolar MVDC distribution system, specifically under pole-to-pole (PTP) fault. The branch currents in the MVDC network are computed using mathematical modeling to provide critical data for fault current estimation. This data is utilized to train an artificial neural network (ANN) algorithm for fault location estimation, with 80% of the data dedicated to training, 10% for testing, and 10% for validation. The ANN achieves a RMSE of 0.8090, MAE of 0.3896, and R² of 0.9912, significantly outperforming support vector regression (SVR). Further analysis, including noise and parameter deviations, shows that the ANN maintains high accuracy, with R² values above 0.97. This confirms the robustness of the proposed method for real-world applications. These results confirm that the proposed ANN-based method offers an accurate and efficient fault location approach for MVDC networks, aiding the development of advanced protection schemes to enhance system reliability and safety.

키워드