This paper focuses on diagnosis of voltage events collected in power distribution networks. Fault networks, induction motor starting, transformer energization and capacitor bank switching cause voltage events. A characterization of voltage events using attributes directly extracted from the voltage and current waveforms is done in this paper. The used attributes are highly related with the event root-cause as well as the relative location of the event source with respect to the measurement point. The relevance of each attribute has been assessed applying a statistical analysis of variance (MANOVA). The most relevant attributes have been used as input to rule-extraction algorithms in order to extract classification rules. The results were obtained using 484 real-world and 38 synthetic voltage events.