Methods of Early Short-Circuit Detection for Low-Voltage Systems

Short-circuits in electrical networks imply extreme mechanical and thermal stresses to loads, systems, and protection devices. The paper discusses new approaches for low-voltage system protection. The three-dimensional locus curves criterion will be pointed out as one opportunity for short-circuit detection. Regression analysis, used to predict the amplitude of a sinusoidal function and therefore the prospective short-circuit current, provides another possible detection algorithm. Also a method based on combined three-phase quantities, e.g. phase currents, has been studied for network failure identification. A comparison and benchmark of the above algorithms for utilization in low-voltage networks has been presented. Moreover, disturbances from typical loads must be considered carefully to avoid nuisance tripping in addition to the aspired rapid failure detection. Locus curve methods increase hardware complexity against both short trip times and very good nuisance tripping avoidance. Instantaneous methods have superior nuisance avoidance and minimal hardware complexity, trading this off against longer detection times. Other possible techniques are the analysis of short-circuit depend voltage drops in the network or wavelet methods interpreting the network currents via dyadic filter banks of a multiresolution analysis.