In this thesis, new schemes for transformer differential protection are proposed. These schemes use different characteristics of the differential currents waveforms under internal fault and magnetizing inrush current conditions. The schemes are based on choosing an appropriate feature of the waveform and monitoring it during the post-disturbance instants. For this purpose, the differential current is considered as a stochastic signal, and two algorithms based on stochastic analysis of differential currents are proposed. The first algorithm converts the main difference between internal fault and magnetizing inrush current into a quantitative description by using the skewness concept of frequency curves. The second scheme introduces the statistical U test method to distinguish the internal faults from the inrush currents in power transformers. Since these new methods are based on a statistical approach rather than some deterministic index, more reliable and accurate judgment is obtained. Testing results indicate that new methods can identify inrush currents promptly and reliably. Key words Transformer differential protection, inrush current, fault current, frequency curves analysis, statistical tests.
