摘要
等值附盐密度是确定污秽等级和绘制电网污区分布图的主要依据,但是,它易受测量用水量的影响,且测量只能在停电状态下进行。通过对3种常用悬式绝缘子进行人工污秽试验,采用BP人工神经网络的方法,建立了以泄漏电流最大值、泄漏电流5个脉冲主成分、环境湿度、温度等8个变量作为输入参数,等值附盐密度作为输出参数的智能预测模型。使用Levenberg-Marquardt快速学习算法对建立的神经网络进行训练。其试验数据验证了该方法的可行性。
The equivalent salt deposit density （ESDD） is the basis of defining pollution classes and mapping pollution areas. The method is, however, usually influenced by the water quantity for measurement, and can only be applied with power off. Artificial pollution tests were carried out on three kinds of typical suspension insulators in this paper. A BP Artificial Neural Network（ANN） was used to establish the intelligent prediction model. The maximum leakage current, five main components of leakage current, humidity and temperature, etc were chosen as eight input variables, and the ESDD was chosen as one output variable. The ANN was trained by the Levenberg-Marquardt fast training algorithm. The feasibility of the method was proved by tests in laboratories and fields.
出处
《绝缘材料》
CAS
2008年第4期58-61,共4页
Insulating Materials
基金
国家自然科学基金资助项目(50377020)
关键词
绝缘子
泄漏电流
人工神经网络
等值附盐密度
insulator
leakage current
artificial neural network
equivalent salt deposit density（ESDD）
