Date of Award


Document Type

Open Access Dissertation


Electrical Engineering


College of Engineering and Computing

First Advisor

Roger A. Dougal


DC power distribution systems (or so called DC microgrids) are of wide interest for various power applications due to their advantages over traditional AC power distribution systems with respect to power density and power distribution efficiency. On the other hand, short-circuit faults present formidable hazards in these systems. It is difficult to extinguish these arc faults via conventional circuit breakers due to the lack of natural zero-crossing of DC current. Also the DC breakers are usually more bulky and costly. Today, fault protection in these systems – to the extent that it exists – relies on over-current time-out limits in power converters or on special circuit breakers that are tripped via over-current or distance relays and that therefore depend critically on a data network. More-robust communication-independent, fully-distributed schemes are needed.

In this dissertation, we address the problems for fault protection in DC microgrids, and define an approach Local Information Based Fault Protection (LIFP) for robust protection against short-circuit faults that does not rely on microgrid-wide communications. Builds on work of Pietro Cairoli, we show how each entity connected to the dc bus, including current-limiting power converters and non-load-breaking disconnect switches, can autonomously detect, identify, and appropriately react to the presence of a short-circuit arcing fault based only on its own local observations of voltage and current. Successful implementation of such an approach can eliminate the need for dc breakers or fuses. Such an approach can rapidly detect a fault, shut down power injection to the bus, isolate the fault, then re-energize and return the bus to service. The entire process can occur in milliseconds and thus can be transparent to load systems that contain small energy buffers.

For MVDC power systems, we extend the coverage of LIFP to arc faults (with arc impedance up to 4 Ω) under varying load conditions (1 pu to 2 pu). The effective resistance of an arc can sometimes be large compared to that of the bus cables, and the arc resistance can vary randomly in time with large bandwidth; these characteristics complicate implementation of the LIFP method. Therefore, the characteristics of arc faults in DC systems are investigated, and the time-average resistance of DC arcs was represented via the Paukert equations, with the coefficients fitted to experimental data from DC converter-fed arcs. We describe the system design constraints and how moving average filters and dynamically-coordinated tripping thresholds can overcome these problems, and then we report the effectiveness of applying the method for a reference system over a wide range of system parameters (e.g. cable size and length), operating conditions (e.g. system current), and fault conditions (arc location, arc length).

In order to validate LIFP, the MATLAB-SIMULINK model of a representative multi-terminal MVDC system was developed. The effectiveness of the method was evaluated by applying arc faults, one at a time, to many locations. The apparent resistance parameter (V/I ratio) was computed for each controllable entity, for each fault occurrence, and evaluated to determine whether appropriate protection action was taken. Also due to the assumption that the current ramping rate (didt⁄) of power converters during any load change is limited to be less than 110A/ms, it is used for differentiating load variations (didt⁄klimit⁄). Results demonstrate that there are no cases where a fault was not disconnected from the system and only a few cases where loads lost power when they optimally should not have. For load variation events with a ramping rate up to 50A/ms, LIFP can successfully identify the incident and initiate adjustment of tripping thresholds within 4.5ms.