Classification of short duration faults (voltage sags) in transmission and distribution power systems

Two techniques for voltage sags characterisation and classification have been integrated. The objective is to assist monitoring systems in order to improve automatic recognition of faults. The abstraction of significant information (temporal and phasorial) is proposed to model faults based on simple descriptors instead of trying to obtain analytical models. A classification of such fault based on phasorial analysis is compared with the results obtained using a learning algorithm that allows an automated and unsupervised classification. Voltage sags waveforms are gathered in a 25kV substation. The goal is to locate the origin of such fault registered in the substation (transmission or distribution).