The proposed contribution describes a method for the detection and classification of harmonic transients, as they occur in electrical networks. The method uses, as bases, a special class of wavelet-packets, i. e. smooth biorthogonal sine and cosine-packets in order to find the best ’’Shannon-basis“. The algorithm is based on the minimization of Shannon‘s function for the decomposed wanted signal over the biorthogonal analyzing functions. The proposed method is validated by using a real application, signals from rail systems. One shows the fastness and the robustness of the algorithm and thus its real time applicability. In particular, one
detects and classifies inrush currents.