This paper presents a new technique for inrush current detection of power transformers. The method is based on wavelet analysis of power system frequency which is estimated by Prony method. The Prony method is utilized to estimate power system frequency with three consecutive samples rapidly. After that, Daubechies wavelet family uses the estimated frequency to investigate the unbalanced conditions in power system. Meanwhile, a new time based criterion is introduced to determine internal fault currents from inrush currents by using the changes of estimated power system frequency. Extended simulations for various faults and inrush conditions demonstrate precise operation of the proposed algorithm also in CT saturation. Keywords -Power transformer protection, inrush current, internal fault, wavelet transform, frequency estimation, Prony method.
I. INTRODUCTIONStability of transformer's protection scheme is one of the critical issues for power system operation performance [1]. One of the fast and reliable protection schemes for power transformer is based on differential current which is called differential protection. Unfortunately, this scheme maloperates in some cases like inrush condition or external fault, which leads to current transformer (CT) saturation. The phenomenon of transformer inrush current has been widely discussed in many papers. In 1904, Merz and Price developed the first approach for differential protection of transformers. The advantage of their proposed scheme was fast recognition and after that the technique has been extensively applied [2]. In order to stabilize the differential protection against maloperation in the event of external faults, the "differential current" tripping parameter is standardized to a stabilization parameter. As one of the other methods, wave shape based recognition methods are divided among those methods that are based on the identification of the separation of different current peaks [3,4]. Other methods for discriminating between internal fault currents and inrush currents are based on direct recognition of the differential current shape. A common well-known principle is discussed in [5]. Ref [6] discusses the length of the time intervals during which the differential current is near zero. Nevertheless, the length of the time interval of inrush currents is decreased in current transformer saturation condition. Some methods use DC offset or asymmetry in the differential current [7][8][9][10][11]. References [12, 13] present wavelet based method for discrimination between inrush currents and internal fault currents, but it does not work in all conditions. Moreover, some of the other methods extract second harmonic of inrush current, but second harmonic is significantly decreased due to the improvement of core steel in modern power transformers [14]. Meanwhile, a large second harmonic component can be introduced in internal fault of transformer in many cases. So, 2 nd harmonic based methods cannot be a final solution to this goal. Proposed method in [15] uses equival...