Automated identification of lightning stroke induced events using signature analysis
MetadataShow full item record
In the recent past, lightning stroke and tree contact triggered disturbances have been the precursors of a number of cascaded power outages in utility transmission and distribution networks. These outages have been responsible for causing significant monetary losses to the industry and the society and damaging installed power equipment as well. Cascaded outages can ideally be prevented if the triggering cause can be detected automatically and in a timely fashion. This report proposes the construction of an expert rule-based event identification system that can be used from a remote monitoring facility to analyze the disturbance events captured by power quality monitors installed at various locations in a distribution network. This event identification system warns its users at the remote monitoring site regarding the initiating cause (lightning stroke or tree contact) of the transient being analyzed. Distinguishing lightning stroke induced events from the tree contact triggered ones automatically, is a highly challenging task. It not only needs extensive knowledge of the phenomena underlying a disturbance but also requires efficient waveform analysis tools to identify and segregate the time dependant and frequency dependant features in the disturbance. Unlike the previous approaches to detect power quality events based on time domain analysis or artificial neural network based waveform classifiers, the event identification system developed in this report utilizes waveform analysis in the wavelet domain as it basis for isolating the disturbance from the nominal power frequency signal in a waveform. This isolation of the disturbance allows the event identification system, thus developed, to examine only the characteristic features of the disturbance which are also representatives of the triggering phenomena underlying it. This event identification system utilizes a decision making algorithm to compare the features extracted from a disturbance with the standard values generated by lightning stroke and tree contact triggered disturbances, in order to identify the triggering phenomena. These standard values are obtained by analyzing disturbances known to be triggered by lightning strokes or tree contacts and are stored in the event identification system in the form of a knowledge database consisting of if-then rules. The prototype system thus built is tested on 72 lightning stroke induced events, 86 tree contact triggered events and 88 unknown events. The accuracy rate of the event identification system in detecting lightning strike events is nearly 93%. But due to certain characteristic limitations of tree contact triggered disturbances, the event identification system is not as efficient in detecting such events, leaving room for further work in this area. This system is expected to help power quality engineers identify the lightning strike events and mitigate the monetary and power equipment losses at stake.