Proposição de uma nova técnica para detecção de faltas de alta impedância em redes de média tensão

Abstract

Power distribution systems are subject to the action of time, people and natural phenomena. These actions can cause the interruption of power supply by the action of the protection system, whether by accident, lightning, by defects in the equipment, among other factors. Conventional protections operate whenever the system detects an overcurrent which may be caused by a short circuit, by current leakage in the insulating or overloaded system. But there is a situation in which these devices do not work. This occurs when a feeder of the primary network breaks down and comes in contact with any surface that has high electrical resistance. This event is called High Impedance Fault (HIF). Under these conditions, the resulting fault current, depending on the terrain, can reach values close to the nominal ones, causing the non action of protection devices. This problem is of great importance, since the fallen feeder remains energized, endangering the lives of people and animals that may go on the local where the conductor remains fallen. Therefore, many studieshave been made to develop methodologies that can detect the HIFs. However, no enough robust method exists that is capable of detecting high impedance faults in all possible operating conditions. Furthermore, most existing methods are not economically viable for large scale implementation. Therefore, this paper proposes a new technique for detecting HIFs using only the values of current sampled at the power substation. This technique has a transient detector and is able to classify a HIF through the feature of buildup of the current waveform, which is characteristic of this type of fault. The detection of a transiente is made according to the calculation of the Euclidean distance between two consecutive points in the plan formed by the current signal samples and their respective values of the first-difference function. Since this parameter undergoes a considerable increase when a disturbance on the network occurs, an event or transient is characterized whenever the value of those distances becomes greater than a threshold value, according to the statistics functions calculated for a data window containing past values of distances. After the occurrence of a transient in the network is verified, it begins the process of high impedance fault classification. An HIF is classified by checking the feature of buildup in the current signal, which remains for a few signal cycles. The proposed algorithm was tested for signals obatined by simulations in the EMTP/ATP, using computer models of HIFs simulation proposed in the literature. The algorithm was implemented in MATLAB R○. The results show that the proposed technique is robust and effective for the detection of HIFs, since it correctly classified the occurrence of high impedance faults in all cases analyzed.