Abstract
This project aims to develop an intelligent system that performs, on an integrated way, the preventive control (proactive system) and the failure diagnosis in electric power distribution systems. It is a procedure to identify, classify and localize critical situations, in an incipient stage, of failures or precursors of failures that can potentially cause damage to system's components and, particularly, the interruption of electric power supply to consumers. This intelligent system is based on wavelet transform, Dempster-Shafer theory of evidence and artificial neural networks, in special the ART (Adaptive resonance Theory) family architecture, i.e., the Fuzzy ARTMAP. Due to the stability and plasticity characteristics, this architecture enables the introduction of the continuous training module, which allows the knowledge extraction without the need to restart the training process when a new training pattern is included, unlike what happens in most neural networks. Thus, it is possible to use a reduced set of patterns on the training phase and, as the analyses are performed, the extraction of the knowledge is continuous, i.e., a system that seeks to improve over time. The system will introduce the main disorders characterized by voltage disturbances and high impedance faults. New types of abnormalities can be incorporated with research's development. Therefore, it deals with a methodology that can permanently incorporate new information to become an efficient inference system.
|