| Summary: | 碩士 === 國立高雄應用科技大學 === 電機工程系 === 99 === Abstract
Zinc-oxide arrester is an important over-voltage protection device in power system because its performance has great influence on the safe operation of electrical equipment. The nonlinear resistive element of Zinc- oxide arrester will be gradually aging under long term operation, resulting in an increase in leakage current. When the aging evolves to a certain condition, a surge caused by lighting or other reasons will result in the thermal collapse of the arrester, and the over-voltage protection functions of the arrester will be lost. According to IEC60099-5 standard, resistive component and third harmonic in leakage current can be used as an indicator of aging and the traditional online monitoring of surge arrester is manually done by instruments such as surge counter and leakage current meter; but, the measuring process is inefficient and prone to cause errors. In this paper, different neural network methods are used for the estimation of arrester aging, and the results show that these algorithms are very effective. Especially for the support-vector-machine method, the accurate rate is up to 100%.
|
|---|
