Fault Location in DC Microgrids using Traveling Waves

In DC power systems, rapid fault location is crucial for maintaining reliable operation, particularly with the prevalence of DC-DC converters. This study investigates fault location techniques in DC systems utilizing Traveling Waves (TWs). Following data normalization, multi-resolution analysis employs discrete wavelet transform to capture high-frequency patterns of TW's wavelet coefficients. Parseval's theorem is utilized to quantify the energy of these coefficients. First, a curve-fitting technique is employed to estimate fault locations in DC microgrids. Then, two transfer learning approaches are proposed: first approach integrates Parseval energy curves into a Gaussian process estimator, while second employs feedforward neural network for fault prediction. Hardware implementation of TW protection device is also explored, involving real-world testing and validation in the Emera Technologies Kirtland Airforce Base low-voltage DC microgrid. Through experimental validation and field tests, the effectiveness of the proposed methodologies in achieving fast and accurate fault location in DC power systems is demonstrated.