극저온 환경에서의 부분방전 신호 검출 및 뉴럴 네트워크를 이용한 패턴인식률 비교

Abstract

The application of the superconducting power apparatus is now being considered as one of the promising tool for enlarging the limited transmission capacity of the traditional electric power apparatus due to the several technical advantages such as reduced size, weight, high efficiency and so on. Therefore, since more than two decades, many research institutes try to improve performance by carrying out experimental investigations related to the reliability of the apparatus under cryogenic temperature. One of them is Partial discharge (PD) detection which is considered as the indication of the insulation state of the apparatus, however, very few reports have been reported based on the results obtained under cryogenic temperature. In this work, 3 different types of artificial defects are put into Liquid Nitrogen in order to produce PD under AC applied voltage: protrusion, floating electrode, and turn to turn. PD signals are detected by use of our specially designed sensor and then its pattern recognition is made based on PRPDA (Phase Resolved PD Analysis) and CAPD (Chaos Analysis of Partial Discharge). Regarding the related recognition rate, NN (Neural Networks) is employed for learning process. Moreover, other patterns from the unknown defects are also put into network for its comparison. On the other hand, difference in recognition rate depending on three methods of NN has been noticed enabling us to deduce their related recognition rate.