Abstarct: Differential relays are main and initial protections of power transformer. Therefore fast and accurate operation of these facts has special importance. In some cases incorrect operation of the differential relays is occurred, that is because of inrush currents, saturation of current transformers, transformer over excitation, and differential current due to unbalanced of current transformer ratio through tapchanger operation. Among them, the impact of inrush current is the most important reason of incorrect operation of differential relay. As a result many studies have been carried out to discriminate between fault and inrush current. In this thesis, a new method for discriminating between fault and inrush current is proposed. In the proposed method, a signal processing tool named as wavelet transform is used. The proposed algorithm is baxsed on the extracted features of frequency and time analysis. The wavelet analysis results in time varient frequency components at different frequency levels. The performed investigations in this thesis on the energy distribution of these components and at the different frequency levels show the energy difference between inrush and internal fault current in the frequency levels. This difference is used to discriminate the fault from inrush currents. In order to form a pattern, and use of it in the proposed algorithm, a statistical measure known as correlation factor is employed. The correlation factor is used to express the relationship between the energies of wavelet transform components at the different frequency levels. In this method a criterion is created and used to block the differential relay in the case of inrush current. Simulation results and experimental labratoary test verify the accuracy, rapidity and reliability of the proposed method.