<p>This paper presents a new method for the characterization and diagnosis of electrical disturbances caused by fuses operation in the electrical distribution systems. A set of descriptors is proposed in order to quantify the typical features of the distortions caused by operation of expulsion and current limiting fuses. A multivariate statistical analysis is performed to select the descriptors with the best profiles qualifiers and the optimal decision thresholds are selected through of machine learning algorithms. Voltage and current signals of the fuses operation are obtained from the ATP-EMTP simulation, as well as some real signals, to be all used in the validation of the new proposed algorithm, obtaining optimal performance and efficiency results. The algorithm was implemented in Matlab and the computational requirements are minimal.</p>