EPPEI Power Seminar #3

Presenter:	Slavko Vasilic Ph.D. Student (advisor: Dr. Kezunovic)
Topic:	Neural-Fuzzy Pattern Recognition Algorithm for Classifying the Events in Power System Networks
Time:	April 26th, Friday at 2:00pm
Location:	119D ZEC

ABSTRACT: This study introduces artificial intelligence based technique for protective relaying of transmission lines in power system networks. An advanced pattern recognition algorithm for classifying the transmission line faults, based on combined use of neural network and fuzzy logic is proposed. Designed algorithm will be used in the efforts aimed at replacing distance relays with new relays not having traditional setting. The new concept is based on special type of neural network, ideally suited for classifying large and diverse set of input data. The approach utilizes adaptive, self-organized neural network where the prevailing power system conditions are taken into account through the learning mechanism. The protection algorithm is based directly on local measurements (sampled currents and voltages of the three transmission line phases) without extraction of traditional features like impedance or phasor computing. The new classification approach has to reliably conclude, in a very short time (one cycle or less), whether, where and which type of the fault occurs under a variety of time-varying operating conditions. Training of the network is based on combined use of unsupervised and supervised learning techniques. Various procedures for preprocessing neural network input signals are implemented and their influence on the algorithm classification capability is investigated. During algorithm evaluation (testing phase), fuzzy decision rule is defined and applied on neural network outputs to improve algorithm selectivity for a variety of test patterns representing real events not seen during training. Presented approach allows implementation of a very fast and robust fault detection and classification algorithm.

Presentation slides