Fault location technique for multi-terminal DC-microgrid based Gaussian process regression

This study presents an effective fault location technique for a multiterminal DC microgrid. Zonal fault detection is first completed by intelligent protection devices that capture voltages and currents, and then the resistance at both cable terminals is estimated. The polarities of the estimated front-end and remote-end resistances are compared by their respective intelligent protection devices to examine the fault. If both polarities are negative, a cable fault has occurred. If the voltages across two small inductors inserted around each junction have opposite polarities, a junction fault is identified. After identifying the faulted line, the proposed Gaussian process regression model is used to determine the fault distance based on one-sided data. The proposed algorithm shows considerable enhancements in terms of detection accuracy and time response. Besides, this technique can successfully identify a range of fault scenarios, such as close-in, mid-point, and remote faults, even amidst noise and under different operational conditions regardless of fault resistance. Furthermore, the method is capable of determining fault distance within 2 ms, achieving an error margin of less than 1%. Extensive simulations and experimental investigations reveal the effectiveness and high accuracy of the proposed approach in detecting and locating various faults under a wide range of fault resistances.