Search for: All records

In this paper we explore the problem of series arc fault detection and localization on dc microgrids. Through a statistical model of the microgrid obtained by nodal equation, the injection currents are modeled as a random vector whose distribution depends on the nodal voltages and the admittance matrix. A series arc fault causes a change in the admittance matrix, which further leads to a change in the data generating distribution of injection currents. The goal is to detect and localize faults on different lines in a timely fashion subject to false alarm constraints. The model is formulated as a quickest change detection problem, and the classical Cumulative Sum algorithm (CUSUM) is employed. The proposed framework is tested on a dc microgrid with active (constant power) loads. Furthermore, a case considering fault detection in the presence of an internal node is presented. Finally, we present an experimental result on a four node dc microgrid to verify the practical application of our approach. 

In this paper, a detection and localization technique based on dual State and Parameter Estimation (SE and PE respectively) for series dc arc faults is presented. Detection of series arc faults in dc microgrids is challenging due to its low fault current. By using the available set of sensor measurement data over a period of time, a Least Squares (LS) based SE algorithm estimates the dc microgrid's bus voltages and injection currents. Kalman Filter (KF) is then used to estimate the line conductances in the network, which are used to detect and localize (with respect to the faulted line) the series arc fault. Simulation results are presented with different case studies to demonstrate the robustness of the algorithm to normal operating conditions and different number and placement of sensors. Finally, Control Hardware in the Loop (CHIL) results are shown.