Reconocimiento de Patrones para la Protección de Líneas de Transmisión ante Descargas Atmosféricas. (Pattern recognition for Lightning Stroke Transmission Line Protection)

MORALES, JOHN

Universidad Nacional de San Juan

Año: 2015 p. 244

978-950-43-0639-9

It is well known that Electric Power Systems (EPS) have been a very importantpart of modern day life, whose main objective is supply electric powermaintaining a very high level of continuity of service. However, abnormal orintolerable conditions can be present especially on Transmission Lines (TLs),which can produce faults. In fact, TLs are permanently under adverse conditionsespecially lightning strokes.At the moment, lightning stroke is an important disturbing phenomenon for TLs.If protection relays are operated erroneously, it can cause destruction ofmaterials, constitutes danger to persons and others. On the other hand, recordsof trip fault on TLs reported by different organizations indicate that at least 60percent are caused by lightning strokes, and in some extreme cases theproportion is close to 90 percent. Accordingly, lightning strokes have been theroot cause of short circuit on TLs and the most important cause of mal-operationof transmission line protection relays.On the other hand, the elimination or total reduction of lightning strokes on EPSis very difficult. Accordingly, it is impossible to avoid their consequences onTLs. Therefore, to analyze this type of phenomenon and correctly classify theirsignals is crucial for the proper protection relays performance.As regards to the protection function, the basic principle of transmission lineprotection has not changed for many years. Both impedance estimationprotection and Transient Based Protection (TBP) have become the most usedprotection relays for the lightning strokes classification. However, it is clear thatcurrently those techniques suffer from drawbacks, which not only cannotcorrectly classify lightning strokes with speed and precision balanced but alsocannot correctly classify lightning strokes with and without fault.Different approaches based particularly on Wavelet Transform (WT) have beenreported to analyze this phenomenon. However, algorithms based on thesetechniques depend on the decomposition level and election of the motherwavelet. Furthermore, previous researches analyze specific conditions oflightning strokes, omitting a wide range of essential parameters for the lightningperformance. In contrast to those traditional techniques, in this thesis work, anovel methodology based on Principal Component Analysis (PCA) for thepatterns extraction of lightning strokes on TLs is presented. On this context,different patterns of transient components generated after the lightning strokehits on transmission lines are extracted. So, extraction and classification oflightning strokes patterns is developed instead of traditional signal processingtechniques, which obtain the knowledge of lightning strokes indirectly.On the other hand, at the moment there is not any developed methodology toobtain patterns, which can be applied to classify transient signals caused byatmospheric discharges on TLs. Hence, the knowledge developed in thisresearch could improve the protection relays performance; reducing theoperation time of traditional algorithms.Based on those different chapters presented in this thesis, different algorithmsfor detecting transient signals and for the identification of switching operationsare presented. Hence, the methodology is strongly recommended to analyze notonly lightning stroke signals but also different signals of switching operations.Furthermore, the viability of the methodology is demonstrated. Thus, acomparative evaluation between PCA and WT is developed, presenting the mostimportant advantages and disadvantages of both techniques.Finally, the author had the opportunity of applying the methodology presentedin this thesis work in order to analyze and study the distribution feeders?performance against lightning. It currently is analyzed by the applicant to thePh.D Eng. Roberto J. Cabral at Federal University of Rio Grande do Sul, PortoAlegre, Brazil.