基于函数逼近能力的母线保护的研究及仿真  

Research and simulation of bus protection with function approximation ability

作　　者：詹红霞[1] 罗建[2] 雷霞[1] 

机构地区：[1]西华大学电气信息学院,四川成都610039 [2]重庆大学高电压工程与电气新技术重点实验室,重庆400044

出　　处：《继电器》2006年第16期12-16,21,共6页Relay

摘　　要：ANN分类能力的ANN保护方法需要大量故障样本,但由于完整的故障样本的获取不易,提出了基于人工神经网络(ANN)函数逼近能力的ANN母线保护方法。函数逼近能力是ANN具有的重要性能之一,依据ANN具有的函数逼近能力,可用ANN模型去替代一个确定的物理对象。母线保护的物理对象是一个输入输出关系确定的函数对象,可用一个ANN模型去替代,或用一个ANN数学模型去逼近母线保护物理对象的输入输出之间的函数关系。通过这个在无故障运行时学习训练出来的母线保护对象的ANN数学模型,就能判断区分母线保护对象的区内和区外故障。For a long time, the application of ANN to relay protection is based on classification ability. Enough fault samples are crucial for the performance of the protection, but limited sample data can be actually available for the training of ANN model. In order to overcome the drawback, bus protection based on ANN model with function approximation ability is presented in this paper. Function approximation is one of the most important ability of ANN, a function object can be replaced by an ANN model with function approximation ability. Physical object of bus protection is a function with certain relation between inputs and outputs, which can be replaced by an ANN model, i.e. can be approximated by an ANN mathematical model. Based on the ANN model trained under normal bus operation conditions, the inner or outer fault can be distinguished successfully.

关 键 词：母线保护 人工神经网络 函数逼近能力 仿真 

分 类 号：TM773[电气工程—电力系统及自动化]