基于模糊输出BP神经网络的绝缘子等值盐密预测  

ESDD Prediction of Insulator Based on Fuzzy Math & BP Artificial Nueral Network

作　　者：何相佑[1] 向凤红[1] 忽建蕊[2] 

机构地区：[1]昆明理工大学 [2]昆明理工大学楚雄应用技术学院,云南楚雄675000

出　　处：《陕西电力》2008年第5期1-4,共4页Shanxi Electric Power

基　　金：国家自然科学基金(50377020)

摘　　要：等值盐密法是衡量绝缘子污秽状态常用的方法,但是,它却不能实时地对污秽状态进行评估。首先对3种常用悬式绝缘子进行人工污秽试验.利用泄漏电流测量系统记录其在运行电压作用下、不同等值盐密、不同相对湿度时的泄漏电流波形并对此进行分析;采用神经网络和模糊数学结合的方法,建立了不同湿度下泄漏电流和等值盐密的关系。选取了具有模糊输出的多层BP神经网络,使用Levenberg—Marquardt快速学习算法对存文建立的模糊输出神经网络进行了训练。最后,利用部分试验数据进行验证。结果表明,使用该方法能够较准确地预测绝缘子的等值盐密。The method of ESDD is usually used to assess the condition of soilage on insulator.Howerver, it cannot assess on real time. Artificial pollution tests were carried out on three kinds of typical suspension insulators in this paper. Waveforms of leakage current under operating voltage, different relative humidity and different ESDD were recorded by a leakage current monitoring system, and the data was analyzed.A method based on the combination of fuzzy math theory and artificial neural network （ANN） was proposed in order to establish the relationship between leakage currents with different relative humidity and ESDD. The many layers of BP ANN are established.The ANN with fuzzy outputs is trained by the Levenberg-Marquardt fast training algorithm. Verigcation of experimental data shows that the ANN is effective to predict the ESDD of outdoor insulators.

关 键 词：绝缘子 泄漏电流 模糊数学 神经网络 等值盐密 

分 类 号：TM73[电气工程—电力系统及自动化]