Abstarct: Abstract Power transformers are among the most expensive and critical pieces of equipment in power networks, where differential protection is one of the fundamental protection schemes. The main challenges of this protection include distinguishing inrush current, sympathetic inrush current, overexcitation, and external faults from internal faults, as well as differentiating between winding faults with and without inrush current, sympathetic inrush current, and overexcitation. Various methods have been proposed for detecting disturbances in this protection scheme; however, they are not necessarily simple to implement and often do not cover all disturbance types. In this paper, an approach is proposed for detecting the disturbances baxsed on the fast discrete orthonormal Stockwell transform of the differential current signal, utilizing statistical indices, including the mean and kurtosis, as well as the relative maximum absolute of the FDOST coefficients within a specified range. Advantages of the proposed approach include simple implementation and low computational burden. The effectiveness of the approach is evaluated using two power transformers, including a real one in the Iranian power network and the other from EMTP–RV, under all disturbance conditions. Obtained results demonstrate the desirable performance of the approach in detecting all disturbances, achieving an accuracy above 99% with an average processing time of 0.2 ms and a detection time of approximately 13 ms. The comparison results between the proposed approach and the harmonic restraint method indicate that the proposed approach exhibits superior performance in detecting disturbances, both with and without noise. Moreover, the approach exhibits robust performance under noisy conditions and current transformer saturation. Also, the approach has high accuracy in detecting experimental internal faults from inrush current recorded on a laboratory transformer.