Lógica paraconsistente anotada de três valores aplicada em raciocínio baseado em casos para diagnóstico de falta em transformador de potência

Abstract

Considering the operator’s difficulty in classifying and diagnosing the fault, this article presents the results obtained through Three Values Annotated Paraconsistent Logic (3vAPL) in Case-based Reasoning (CBR) in order to identify different types of faults in the Electrical Power Systems. The faults’ classification and identification have the alarms - associated with the triggering of protection relays - as information basis. After the shutting down of the electrical network for protection, it is up to the operator to select the most relevant messages, to extract a conclusion based on the available data and to suitably act towards the reestablishment of the energy. For the work in question, protection is constituted of five relays, being: differential 87, Buchholz 63, restricted earth fault relay 64, phase overcurrent 51, timed and neutral overcurrent, timed 51N. The five relays generate 32 operation possibilities; only five of them are known. The five known operations compose a matrix (5x5) called Knowledge Basis. the functioning of the 3vAPL in CBR uses the Modified Cosine Matching Function (MCMF) to establish a similarity degree (Gcas) between the Knowledge Basis and the other 27 unknown combinations. From the Gcas, it is possible to identify the events of the 27 combinations through the 3vAPL in CBR. Aiming at validation, the results have been compared to those obtained through the Bayes theorem. The tests have been carried out based on the protection scheme of Power Transformers (PT), which are formed of five protection functions. The results show that the proposed model presents a performance that is superior to that of the Generalized Neural Network (GRNN). Finally, the results of this work show that the 3vAPL in CBR, when compared to the GRNN Neural Networks, present a superior result, besides showing a 100% assertiveness when compared to the classification and identification table of faults by the Bayes theorem